path: root/Master/texmf-dist/scripts/fontools/autoinst

#! /usr/bin/env perl

=begin COPYRIGHT

----------------------------------------------------------------------------

    Copyright (C) 2005-2020 Marc Penninga.

    This program is free software; you can redistribute it and/or
    modify it under the terms of the GNU General Public License
    as published by the Free Software Foundation, either version 2
    of the License, or (at your option) any later version.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with this program; if not, write to
        Free Software Foundation, Inc.,
        59 Temple Place,
        Suite 330,
        Boston, MA 02111-1307,
        USA

----------------------------------------------------------------------------

=end COPYRIGHT

=cut

use strict;
use warnings;

use Cwd ();
use File::Path ();
use File::Spec ();
use Getopt::Long ();
use Pod::Usage ();
use POSIX ();

my $VERSION = '20200129';

my ($d, $m, $y) = (localtime time)[3 .. 5];
my $TODAY = sprintf "%04d/%02d/%02d", $y + 1900, $m + 1, $d;

=begin Comment

    Some fontnames contain abbreviated words for width, weight and/or shape;
    we unabbreviate these using the following table.

    To avoid having to have this information in multiple places,
    we also build a reversed table %ABBREV that maps 'full' forms
    to all known abbreviations; we then inject these abbreviations into
    the %NFSS_WEIGHT, %NFSS_WIDTH and %NFSS_SHAPE tables below,
    so that they recognise the abbreviated forms as well as the full ones.

=end Comment

=cut

my %UNABBREVIATE = (
    cmp     =>  'compressed',
    comp    =>  'compressed',
    cond    =>  'condensed',
    demi    =>  'demibold',
    extcond =>  'extracondensed',
    hair    =>  'hairline',
    incline =>  'inclined',
    it      =>  'italic',
    ita     =>  'italic',
    md      =>  'medium',
    slant   =>  'slanted',
    ultra   =>  'ultrablack',
);
my %ABBREV;
while ( my ($k, $v) = each %UNABBREVIATE ) {
    push @{$ABBREV{$v}}, $k;
}
for my $full (keys %ABBREV) {
    push @{$ABBREV{$full}}, $full;
}

=begin Comment

    LaTeX's NFSS contains a number of standard codes for weight and width:
    - weight: ul, el, l, sl, m, sb, b, eb, ub
    - width:  uc, ec, c, sc, m, sx, x, ex, ux

    These codes are not always a perfect match with the weights and widths
    present in a font family; some families (especially many sans serif ones)
    contain more or different weights and widths, and the naming of those
    weights and widths isn't always consistent between font families.
    To handle this situation, we use a two-tiered approach:
    1.  We install all fonts using a "series" name that is the concatenation
        of whatever the font designer has chosen to call the weight and width
        (but in all *lower*case).
    2.  We add "alias" rules to the .fd files that map the standard NFSS codes
        to actual fonts.

    In step 1, we follow NFSS in leaving out any occurrence of
    the word "regular" unless *both* weight and width are Regular;
    in that case, the 'series' attribute becomes "regular".

    The two tables NFSS_WEIGHT and NFSS_WIDTH are used to control step 2.
    It contains several entries of the form

        sc  =>  [ qw( semicondensed narrow ) ],

    This should be read as follows: the NFSS code "sc" is mapped to
    the *first* width on the right hand side present in the current family.

    Please note that the tables contain empty keys instead of "m" for the
    regular weight and width. NFSS actually combines weight and width into
    a single "series" attribute; a weight or width of "m" is left out of
    this combination (unless *both* weight and width are equal to "m"; then
    the series becomes "m", but that's a special case we deal with later on).

    In addition to the mapping of NFSS codes, the two mentioned tables are
    also used in parsing the font's metadata to determine its weight and
    width: any string that occurs on the right hand side is considered a
    possible name to be searched for.

    These tables can be extended to teach autoinst about new weights or
    widths.  Suppose your font family contains a "Hemibold" weight, that
    you want mapped to the "sb" code. Then add the name "hemibold" to
    the right hand side of the "sb" entry in the NFSS_WEIGHT table:

        sb  =>  [ qw( semibold demi demibold medium hemibold ) ],

    In this case, since it's in last position, it's only mapped to "sb"
    if none of the other fonts are present. Put it earlier in the list
    to give it higher priority.

    Note that autoinst converts all metadata to lowercase to avoid
    inconsistent capitalization; so all entries in these tables should
    be *lowercase* as well.

    Technical notes:
    -   We define NFSS_WEIGHT and NFSS_WIDTH first as arrays
        and then as hashtables; this allows us to use the array-variants
        as an *ordered* (by weight/width) list of values.
    -   We exclude the Medium and Regular weights and widths
        from the @WEIGHTS and @WIDTHS arrays to avoid false positives;
        see the comments in the parse_basicinfo() function.

=end Comment

=cut

my @NFSS_WEIGHT = (
    ul  =>  [ qw( ultralight thin 100 hairline eight four two ) ],
    el  =>  [ qw( extralight 200 ) ],
    l   =>  [ qw( light 300 ) ],
    sl  =>  [ qw( semilight blond ) ],
    ''  =>  [ qw( regular text book 400 ) ],
    sb  =>  [ qw( semibold demibold 600 medium 500 ) ],
    b   =>  [ qw( bold 700 ) ],
    eb  =>  [ qw( extrabold 800 ) ],
    ub  =>  [ qw( ultrabold black heavy extrablack ultrablack 900
                  ultraheavy poster super 1000 ) ],
);
my %NFSS_WEIGHT = @NFSS_WEIGHT;
@NFSS_WEIGHT = grep { !ref } @NFSS_WEIGHT;

# Add abbreviated forms, using the %ABBREV table constructed earlier
for my $code (@NFSS_WEIGHT) {
    $NFSS_WEIGHT{$code}
        = [ map { @{ $ABBREV{$_} // [$_] } } @{$NFSS_WEIGHT{$code}} ];
}

my @WEIGHTS = grep { $_ !~ m/ regular | medium /xms }
                   map { @{$_} } values %NFSS_WEIGHT;

my @NFSS_WIDTH = (
    uc  =>  [ qw( ultracondensed extracompressed ultracompressed ) ],
    ec  =>  [ qw( extracondensed compressed compact ) ],
    c   =>  [ qw( condensed ) ],
    sc  =>  [ qw( semicondensed narrow ) ],
    ''  =>  [ qw( regular ) ],
    sx  =>  [ qw( semiextended semiexpanded ) ],
    x   =>  [ qw( extended expanded wide ) ],
    ex  =>  [ ],
    ux  =>  [ ],
);
my %NFSS_WIDTH = @NFSS_WIDTH;
@NFSS_WIDTH = grep { !ref } @NFSS_WIDTH;

# Add abbreviated forms, using the %ABBREV table constructed earlier
for my $code (@NFSS_WIDTH) {
    $NFSS_WIDTH{$code}
        = [ map { @{ $ABBREV{$_} // [$_] } } @{$NFSS_WIDTH{$code}} ];
}

my @WIDTHS = grep { $_ ne 'regular' } map { @{$_} } values %NFSS_WIDTH;

=begin Comment

    The NFSS_SHAPE table maps various shape names to NFSS codes.

    Like in the other NFSS_* tables, entries may be added to teach autoinst
    about new shapes.

=end Comment

=cut

my @NFSS_SHAPE = (
    roman       =>  'n',
    upright     =>  'n',
    italic      =>  'it',
    inclined    =>  'sl',
    oblique     =>  'sl',
    slanted     =>  'sl',
    romani      =>  'n',    # Silentium has two roman shapes, but no italic;
    romanii     =>  'it',   # so we cheat by mapping the second roman to 'it'
);
my %NFSS_SHAPE = @NFSS_SHAPE;
for my $full (keys %ABBREV) {
    if (defined $NFSS_SHAPE{$full}) {
        for my $abbrev ( @{$ABBREV{$full}} ) {
            $NFSS_SHAPE{$abbrev} = $NFSS_SHAPE{$full};
        }
    }
}

=begin Comment

    The %SHAPE table is used in deciding which font shapes
    (normal, small caps, swash or textcomp) to generate.

    Each key in this table names a shape; the corresponding value
    is an anonymous hash with several key/value pairs:
        code    An anonymous hash with three possible keys:
                'n'  -> the NFSS code to use for this variant shape
                        if the 'basic shape' is upright;
                'it' -> the NFSS code to use for this variant shape
                        if the 'basic shape' is italic
                'sl' -> the NFSS code to use for this variant shape
                        if the 'basic shape' is slanted (aka oblique);
                If any entry is missing, the corresponding version
                of this variant shape will not be built.
        reqd    A list of required OpenType features;
                this shape is built if the font supports at least *one*
                of these features.
        nice    A list of optional OpenType features;
                these are used if the font supports them, but don't
                prevent this shape from being built when missing.
        extra   Extra options passed to otftotfm when creating this shape.
        name    A string added to the name of the generated font,
                to make it unique.

    Textcomp is treated as a 'shape' even though it is technically
    an encoding; that is just the easiest way to do things.

=end Comment

=cut

my %SHAPE = (
    normal => {
        code  => { n => 'n', it => 'it', sl => 'sl' },
        reqd  => [ ],
        nice  => [ 'kern', 'liga' ],
        extra => '',
        name  => '',
    },
    smallcaps => {
        code  => { n => 'sc', it => 'scit', sl => 'scsl' },
        reqd  => [ 'smcp' ],
        nice  => [ 'kern', 'liga' ],
        extra => '--unicoding="germandbls =: SSsmall"',
        name  => 'sc',
    },
    # '--feature=aalt' is in 'extra' instead of 'nice' because 'extra'
    # comes last in the command line; else 'aalt' might be overridden
    swash => {
        code  => { n => 'nw', it => 'sw' },
        reqd  => [ 'swsh' ],
        nice  => [ 'kern', 'liga', 'dlig' ],
        extra => '--include-alternates="*.swash" --feature=aalt',
        name  => 'swash',
    },
    textcomp => {
        code  => { n => 'n', it => 'it', sl => 'sl' },
        reqd  => [ ],
        nice  => [ 'onum' ],
        extra => '',
        name  => '',
    },
);

=begin Comment

    The SSUB_SHAPE table is used to generate extra ssub rules in .fd files
    to map missing Slanted shapes to Italic and vice versa.

=end Comment

=cut

my %SSUB_SHAPE = (
    sl      =>  'it',
    scsl    =>  'scit',
    it      =>  'sl',
    scit    =>  'scsl',
);

=begin Comment

    The %STYLE table is used in deciding which figure styles to generate.
    Each figure style (lining, oldstyle, tabular, proportional, superior,
    inferior etc.) becomes a separate font family. We also treat Ornaments
    as a figure style here; that's just the easiest way to handle them.

    Each key in this table names a figure style; the corresponding
    value is an anonymous hash with four key/value pairs:
        reqd    A list of required OpenType features; this style is built
                if the font supports *all* these features.
        nice    A list of optional OpenType features;
                these are used if the font supports them, but don't
                prevent this style from being built when missing.
        extra   Extra options passed to otftotfm when creating this style.
        shape   An anonymous array of 'variant' shapes to build with
                this figure style.

    The 'reqd' and 'nice' subtables for the TLF, LF, TOsF and OsF styles
    are empty; these are filled in at run time, depending on
    which figure style is default for the current font.

    The 'reqd' subtable for the Inf style is also empty; this may be filled
    with 'subs', 'sinf' or 'dnom' depending on the -inferiors options.

=end Comment

=cut

my %STYLE = (
    TLF => {
        reqd   => [ ],
        nice   => [ ],
        extra  => '',
        shapes => [ 'normal', 'smallcaps', 'swash', 'textcomp' ],
    },
    LF => {
        reqd   => [ ],
        nice   => [ ],
        extra  => '',
        shapes => [ 'normal', 'smallcaps', 'swash', 'textcomp' ],
    },
    TOsF => {
        reqd   => [ ],
        nice   => [ ],
        extra  => '',
        shapes => [ 'normal', 'smallcaps', 'swash', 'textcomp' ],
    },
    OsF => {
        reqd   => [ ],
        nice   => [ ],
        extra  => '',
        shapes => [ 'normal', 'smallcaps', 'swash', 'textcomp' ],
    },
    Sup => {
        reqd   => [ 'sups' ],
        nice   => [ ],
        extra  => '--ligkern="* {KL} *"',
        shapes => [ 'normal' ],
    },
    Inf => {
        reqd   => [ ],
        nice   => [ ],
        extra  => '--ligkern="* {KL} *"',
        shapes => [ 'normal' ],
    },
    Numr => {
        reqd   => [ 'numr'],
        nice   => [ ],
        extra  => '--ligkern="* {KL} *"',
        shapes => [ 'normal' ],
    },
    Dnom => {
        reqd   => [ 'dnom' ],
        nice   => [ ],
        extra  => '--ligkern="* {KL} *"',
        shapes => [ 'normal' ],
    },
    Titl => {
        reqd   => [ 'titl' ],
        nice   => [ 'kern', 'liga' ],
        extra  => '',
        shapes => [ 'normal' ],
    },
    Orn => {
        reqd   => [ 'ornm' ],
        nice   => [ ],
        extra  => '--ligkern="* {KL} *"',
        shapes => [ 'normal' ],
    },
);

=begin Comment

    The %STYLE_DEFAULTS table is used to populate the 'reqd' and 'nice'
    subtables of the %STYLE table, depending on the default figure style
    for the current font. This is needed because some font families
    provide only features for the non-default figure styles.

    We therefore make the OpenType features that correspond
    to the default figure style 'nice' (optional), so that this figure style
    is always generated, even without OpenType features in the font;
    the other figure style are only generated when the corresponding
    features are present in the font.

=end Comment

=cut

my %STYLE_DEFAULTS = (
    lnum =>
        sub {
            push @{$STYLE{TLF}{nice}},  'lnum';
            push @{$STYLE{LF}{nice}},   'lnum';
            push @{$STYLE{TOsF}{reqd}}, 'onum';
            push @{$STYLE{OsF}{reqd}},  'onum';
        },
    onum =>
        sub {
            push @{$STYLE{TLF}{reqd}},  'lnum';
            push @{$STYLE{LF}{reqd}},   'lnum';
            push @{$STYLE{TOsF}{nice}}, 'onum';
            push @{$STYLE{OsF}{nice}},  'onum';
        },
    tnum =>
        sub {
            push @{$STYLE{TLF}{nice}},  'tnum';
            push @{$STYLE{TOsF}{nice}}, 'tnum';
            push @{$STYLE{LF}{reqd}},   'pnum';
            push @{$STYLE{OsF}{reqd}},  'pnum';
        },
    pnum =>
        sub {
            push @{$STYLE{TLF}{reqd}},  'tnum';
            push @{$STYLE{TOsF}{reqd}}, 'tnum';
            push @{$STYLE{LF}{nice}},   'pnum';
            push @{$STYLE{OsF}{nice}},  'pnum';
        },
);


############################################################################


sub main {
    print "autoinst, version $VERSION\n";

    ARGV::parse_options();

    my @fonts = map { Fontinfo::parse_fontinfo($_) } @ARGV;

    $ARGV{logfile} ||= sprintf "%s.log", lc $fonts[0]->{family};
    create_logfile(@fonts);
    Fontinfo::assert_unique(@fonts);

    # We can only handle the '-inferiors=auto' option now,
    # since we need to know which inferior figures this font supports;
    # so we have to do the font info parsing first.
    if ($ARGV{inferiors} eq 'auto') {
        Fontinfo::handle_auto_inferiors(@fonts);
    }

    # For each font, figure out the styles, shapes and encodings to generate
    my @worklist
        = cleanup(expand_encodings(expand_shapes(expand_styles(@fonts))));

    # Organize the worklist by family, encoding, style, series and shape
    my %fddata;
    for my $item (@worklist) {
        my $family   = $item->{font}{family};
        my $encoding = $item->{encoding}[1];
        my $style    = $item->{style};
        my $series   = $item->{font}{series};
        my $shape    = $item->{fdshape};
        my $minsize  = $item->{font}{minsize};
        my $maxsize  = $item->{font}{maxsize};

        push @{$fddata{$family}{$encoding}{$style}{$series}{$shape}},
             [ $minsize, $maxsize, $item->{fontname} ];
    }

    # Decide to which weights and widths we'll map the standard NFSS codes
    decide_nfss_mappings(@fonts);

    log_worklist(@worklist);

    if (!$ARGV{dryrun}) {
        # Create the LaTeX support files
        while (my ($fam, $famdata) = each %fddata) {
            LaTeX::write_stylefile($fam, $famdata);
            while (my ($enc, $encdata) = each %$famdata) {
                while (my ($sty, $stydata) = each %$encdata) {
                    LaTeX::write_fdfile($fam, $enc, $sty, $stydata);
                }
            }
        }

        # Create (and execute) commands for otftotfm
        make_commands(@worklist);
    }

    close $ARGV{logfile};
    print "\n";

    return;
}

#-----------------------------------------------------------------------
# Return a list with all unique entries from the input
#-----------------------------------------------------------------------
sub uniq {
    my %seen;
    return grep { !$seen{$_}++ } @_;
}

#-----------------------------------------------------------------------
# Determine which figure styles to create for each font;
# return a list of (font, style) tuples
#-----------------------------------------------------------------------
sub expand_styles {
    return map { my $font = $_;
                 map { { font => $font, style => $_ } }
                     grep { has_all_reqd_features($font, $STYLE{$_}) }
                          keys %STYLE
               }
               @_;
}

#-----------------------------------------------------------------------
# Check if $font has all features mentioned in the 'reqd' subtable of $target
#-----------------------------------------------------------------------
sub has_all_reqd_features {
    my ($font, $target) = @_;

    return 0 unless defined $target;
    for my $prereq (@{$target->{reqd}}) {
        return 0 if !$font->{feature}{$prereq};
    }
    return 1;
}

#-----------------------------------------------------------------------
# Determine which shapes to create for each (font, style) combination;
# return a list of (font, style, shape) tuples
#-----------------------------------------------------------------------
sub expand_shapes {
    return map { my ($font, $style) = @{$_}{qw(font style)};
                 map { { font => $font, style => $style, shape => $_ } }
                     grep { has_some_reqd_shapes($font, $SHAPE{$_}) }
                          @{$STYLE{$style}{shapes}};
               }
               @_;
}

#-----------------------------------------------------------------------
# Check if $font has at least one of the features mentioned
# in the 'reqd' subtable of $target
#-----------------------------------------------------------------------
sub has_some_reqd_shapes {
    my ($font, $target) = @_;

    return 0 unless defined $target->{code}{$font->{basicshape}};
    return 1 unless @{$target->{reqd}};
    return grep { $font->{feature}{$_} } @{$target->{reqd}};
}

#-----------------------------------------------------------------------
# Determine which encodings to use for each (font, style, shape) combi;
# return a list of (font, style, shape, encoding) tuples
#-----------------------------------------------------------------------
sub expand_encodings {
    return
        map { my ($font, $style, $shape) = @{$_}{qw(font style shape)};
              my @encodings
                  = $shape eq 'textcomp' ? (['fontools_ts1', 'ts1'])
                  : $style eq 'Orn'      ? ([Fontinfo::get_orn($font), 'u'])
                  : map { [lc $_, lc $_] } @{$ARGV{encoding}}
                  ;
              map {  { font     => $font,
                       style    => $style,
                       shape    => $shape,
                       encoding => $_,
                     }
                  }
                  @encodings;
            }
            @_;
}

#-----------------------------------------------------------------------
# Add some finishing touches to the work list
#-----------------------------------------------------------------------
sub cleanup {
    my @worklist = @_;

    WORKITEM:
    for my $item (@worklist) {
        my ($font, $style, $shape, $encoding)
            = @{$item}{qw(font style shape encoding)};

        # don't generate smallcaps version of TS1-encoded fonts,
        # as these contain the same glyphs as the regular version
        if ($font->{is_smallcaps}) {
            if ($shape eq 'textcomp') {
                $item = 0;
                next WORKITEM;
            }
            else {
                $shape = 'smallcaps';
            }
        }

        # generate unique name for this font
        $item->{fontname} = join '-', grep { $_ } $font->{name},
                                                  lc $style,
                                                  lc $SHAPE{$shape}{name},
                                                  lc $encoding->[1];

        # look up NFSS code for font's shape
        $item->{fdshape} = $SHAPE{$shape}{code}{$font->{basicshape}};

        # compile list of OpenType features to use with this font
        my %feature = map { ($_ => 1) }
                          grep { $font->{feature}{$_} }
                               ( @{$STYLE{$style}{reqd}},
                                 @{$STYLE{$style}{nice}},
                                 @{$SHAPE{$shape}{reqd}},
                                 @{$SHAPE{$shape}{nice}},
                               );
        if ($feature{lnum} && $feature{onum}) {
            delete $feature{lnum};
        }
        $item->{features} = [ sort keys %feature ];

        # add finishing touches to name and NFSS code of encoding
        $item->{encoding}[0]
            =~ s/\A (ot1|t1|ly1|lgr|t2[abc]|t3|ts3) \z/fontools_$1/xms;
        $item->{encoding}[1] = uc $item->{encoding}[1];

        $item->{cmdline} = make_cmdline($item);
    }

    return grep { $_ } @worklist;
}

#-----------------------------------------------------------------------
# Modify NFSS_WEIGHT and _WIDTH tables: pick one choice for each NFSS code
#-----------------------------------------------------------------------
sub decide_nfss_mappings {
    my ( %weight, %width );
    for my $font (@_) {
        $weight{ $font->{weight} } //= $font->{weight_class};
        $width{ $font->{width} } //= $font->{width_class};
    }

    NFSSWEIGHT:
    for my $nfssweight (@NFSS_WEIGHT) {
        $NFSS_WEIGHT{$nfssweight}
            = [ grep { $weight{$_} } @{$NFSS_WEIGHT{$nfssweight}} ];
    }

    # Some trickery to handle the case where the ul/ub codes are mapped
    # but the el/eb codes are still empty. We try two things:
    # 1.  if there is a Thin (Heavy) weight and this is less extreme
    #     than the weight mapped to ul (ub), we map Thin (Heavy) to ul (ub)
    # 2.  otherwise we move the ul/ub weight to the el/eb position,
    #     unless that weight is the Ultralight/Ultrabold weight
    if ( !$ARGV{el} and !$ARGV{ul} ) {
        if ( @{$NFSS_WEIGHT{ul}} and !@{$NFSS_WEIGHT{el}} ) {
            if ( $weight{thin}
                    and $weight{thin} > $weight{$NFSS_WEIGHT{ul}[0]} ) {
                $NFSS_WEIGHT{el} = [ 'thin', ];
            }
            elsif ( $NFSS_WEIGHT{ul}[0] ne 'ultralight' ) {
                $NFSS_WEIGHT{el} = [ shift @{$NFSS_WEIGHT{ul}} ];
            }
        }
    }
    if ( !$ARGV{eb} and !$ARGV{ub} ) {
        if ( @{$NFSS_WEIGHT{ub}} and !@{$NFSS_WEIGHT{eb}} ) {
            if (    $weight{heavy}
                and $weight{heavy} < $weight{$NFSS_WEIGHT{ub}[0]}
            ) {
                $NFSS_WEIGHT{eb} = [ 'heavy', ]
                    unless @{$NFSS_WEIGHT{b}}
                       and $weight{$NFSS_WEIGHT{b}[0]} > $weight{heavy};
            }
            elsif ( $NFSS_WEIGHT{ub}[0] ne 'ultrabold' ) {
                $NFSS_WEIGHT{eb} = [ shift @{$NFSS_WEIGHT{ub}} ];
            }
        }
    }

    # Special case: if we don't have Regular but we *do* have Medium,
    # move Medium from the "sb" list to the "m" (i.e., Regular) one.
    if ( !@{$NFSS_WEIGHT{''}} ) {
        my $alternate = ( grep { $weight{$_} } qw(medium 500) )[0];
        if ($alternate) {
            $NFSS_WEIGHT{''} = [ $alternate ];
            $NFSS_WEIGHT{sb}
                = [ grep { $_ ne $alternate } @{$NFSS_WEIGHT{sb}} ];
        }
    }

    # Some more trickery to map the sl code to Book or Text (but of course
    # only if sl is empty and Book/Text is lighter than Regular)
    if ( !@{$NFSS_WEIGHT{sl}} ) {
        $NFSS_WEIGHT{sl}
            = [ grep { $weight{$_} < $weight{$NFSS_WEIGHT{''}[0]} }
                     @{$NFSS_WEIGHT{''}}
            ];
    }

    NFSSWIDTH:
    for my $nfsswidth (@NFSS_WIDTH) {
        for my $width ( @{$NFSS_WIDTH{$nfsswidth}} ) {
            if ( $width{$width} ) {
                $NFSS_WIDTH{$nfsswidth} = [ $width ];
                next NFSSWIDTH;
            }
        }
        $NFSS_WIDTH{$nfsswidth} = [];
    }
}

#-----------------------------------------------------------------------
# Log all fonts, styles, encodings etc. we're  going to create
#-----------------------------------------------------------------------
sub create_logfile {
    my @fonts = @_;

    if (-e $ARGV{logfile}) {
        print "[WARNING]   File '$ARGV{logfile}' already exists;\n" .
              "            appending new log data to end.\n";
    }
    open my $LOG, '>>', $ARGV{logfile}
        or die "$0: cannot create $ARGV{logfile}: $!";
    $ARGV{logfile} = $LOG;

    print {$LOG} <<"END_ARGUMENTS";


############################################################################


@{[ POSIX::strftime("[%F %T]", localtime time) ]}  $0, version $VERSION

    "$ARGV{cmdline}"

----------------------------------------------------------------------------

I'm using the following options:

    encoding(s):        @{[ join ', ', @{$ARGV{encoding}} ]}
    NFSS:               $ARGV{nfss} @{[ $ARGV{nfss} eq 'rm' ? '(serif)'
                                      : $ARGV{nfss} eq 'sf' ? '(sanserif)'
                                      : $ARGV{nfss} eq 'tt' ? '(typewriter)'
                                      :                       '(unknown)'
                                    ]}

    lining:             @{[ $ARGV{lining}       ? 'yes' : 'no' ]}
    oldstyle:           @{[ $ARGV{oldstyle}     ? 'yes' : 'no' ]}
    proportional:       @{[ $ARGV{proportional} ? 'yes' : 'no' ]}
    tabular:            @{[ $ARGV{tabular}      ? 'yes' : 'no' ]}
    ts1:                @{[ $SHAPE{textcomp}    ? 'yes' : 'no' ]}
    smallcaps:          @{[ $ARGV{smallcaps}    ? 'yes' : 'no' ]}
    swash:              @{[ $ARGV{swash}        ? 'yes' : 'no' ]}
    titling:            @{[ $ARGV{titling}      ? 'yes' : 'no' ]}
    superiors:          @{[ $ARGV{superiors}    ? 'yes' : 'no' ]}
    inferiors:          $ARGV{inferiors}
    ornaments:          @{[ $ARGV{ornaments}    ? 'yes' : 'no' ]}
    fractions:          @{[ $ARGV{fractions}    ? 'yes' : 'no' ]}
    ligatures:          @{[ $ARGV{ligatures}    ? 'yes' : 'no' ]}

    auto/manual:        @{[ $ARGV{manual}       ? 'manual'  : 'auto' ]}
    target:             $ARGV{target}
    extra:              $ARGV{extra}

    figurekern:         @{[ $ARGV{figurekern}   ? 'keep' : 'remove' ]}
    mergewidths:        @{[ $ARGV{mergewidths}  ? 'yes' : 'no' ]}
    mergeweights:       @{[ $ARGV{mergeweights} ? 'yes' : 'no' ]}
    mergeshapes:        @{[ $ARGV{mergeshapes}  ? 'yes' : 'no' ]}

    nfssweight:         @{[ join q{, }, @{$ARGV{nfssweight}} ]}
    nfsswidth:          @{[ join q{, }, @{$ARGV{nfsswidth}}  ]}

    @{[ $ARGV{dryrun} ? 'DRY RUN' : '' ]}

END_ARGUMENTS

    return if $ARGV{verbose} < 1;

    print {$LOG} '-' x 76 . "\n\nResults of font info parsing:\n";

    for my $font (@fonts) {
        print {$LOG} <<"END_PARSE_FONT";

    $font->{filename}
        Name:       $font->{name}
        Family:     $font->{family}
        Subfamily:  $font->{subfamily}
        Weight:     $font->{weight} ($font->{weight_class})
        Width:      $font->{width} ($font->{width_class})
        Shape:      $font->{shape} @{[ $font->{is_smallcaps}
                                            ? 'smallcaps' : '' ]}
        Size:       $font->{minsize}-$font->{maxsize}
        Features:   @{[ join ', ', sort keys %{$font->{feature}} ]}
END_PARSE_FONT
    }
}

#-----------------------------------------------------------------------
# Log all fonts, styles, encodings etc. we're  going to create
#-----------------------------------------------------------------------
sub log_worklist {
    my @worklist
        = sort { $a->{font}{filename} cmp $b->{font}{filename}
                    || $a->{encoding}[1] cmp $b->{encoding}[1]
                    || $a->{style} cmp $b->{style}
               } @_;

    my $LOG = $ARGV{logfile};

    print {$LOG} "\n" . '-' x 76 . "\n\nNFSS mappings:\n\n";
    for my $weight (@NFSS_WEIGHT) {
        printf {$LOG} "    %-3s =>  %s\n",
                $weight || 'm', $NFSS_WEIGHT{$weight}[0] || '';
    }
    printf {$LOG} "\n";
    for my $width (@NFSS_WIDTH) {
        printf {$LOG} "    %-3s =>  %s\n",
                $width || 'm', $NFSS_WIDTH{$width}[0] || '';
    }
    printf {$LOG} "\n";

    return if $ARGV{verbose} < 1;

    my ($prevfn, $prevsty, $prevenc) = ('') x 3;
    my @cmds;
    for my $item (@worklist) {
        if ($prevfn ne $item->{font}{filename}) {
            print {$LOG} <<"END_FONTINFO";

    ------------------------------------------------------------------------

    $item->{font}{filename}

        Generating these encodings, figure styles and shapes:

        ENC     STYLE   SHAPE   FEATURES USED
END_FONTINFO
        }
        if ($prevenc ne $item->{encoding}[1]
            || $prevsty ne $item->{style}) {
            print {$LOG} "\n";
        }
        printf {$LOG} "        %-3s     %-4s    %-4s    %s\n",
                      $item->{encoding}[1],
                      $item->{style},
                      $item->{fdshape},
                      join(', ', @{$item->{features}}),
                      ;
        $prevfn  = $item->{font}{filename};
        $prevsty = $item->{style};
        $prevenc = $item->{encoding}[1];

        push @cmds, $item->{cmdline};
    }

    return if $ARGV{verbose} < 2;
    print {$LOG} "\n\n";
    print {$LOG} join "\n\n", @cmds;
    print {$LOG} "\n";
}

#-----------------------------------------------------------------------
# Generate all otftotfm commands, and either save or execute them
#-----------------------------------------------------------------------
sub make_commands {
    my @worklist = @_;

    my @commands = map { $_->{cmdline} } @worklist;

    # make sure the last command *does* call updmap
    $commands[-1] =~ s/--no-updmap//xms if $ARGV{updmap};

    if ($ARGV{manual}) {
        open my $BAT, '>', 'autoinst.bat'
            or die "[ERROR]     Can't create 'autoinst.bat': $!";
        print {$BAT} "$_\n" for @commands;
        close $BAT;
    }
    else {
        my $oops = 0;
        $| = 1;     # turn on autoflush, to make a poor man's progress bar
        print "[INFO]      Generating fonts ";
        for my $command (@commands) {
            print '.';
            open my $otftotfm, '-|', "$command 2>&1"
                or die "could not fork(): $!";
            my $msgs = do { local $/; <$otftotfm> };
            close $otftotfm
                or do {
                    warn "\n$command\n\n$msgs\n";
                    print {$ARGV{logfile}} "\n$command\n\n$msgs\n";
                    $oops = 1;
                };
        }
        print "\n";
        $| = 0;
        if ($oops) {
            warn <<"END_OTFTOTFM_WARNING";
[ERROR]     One or more calls to 'otftotfm' returned a non-zero status code;
            please check the messages above and in the log file.
END_OTFTOTFM_WARNING
        }
    }

    return;
}

# The official names for various coding schemes
my %SCHEME = (
    fontools_ot1 => 'TEX TEXT',
    fontools_t1  => 'EXTENDED TEX FONT ENCODING - LATIN',
    fontools_ts1 => 'TEX TEXT COMPANION SYMBOLS 1---TS1',
    fontools_ly1 => 'TEX TYPEWRITER AND WINDOWS ANSI',
    fontools_lgr => 'GREEK FONT ENCODING - LGR',
    fontools_t2a => 'TEX CYRILLIC FONT ENCODING - T2A',
    fontools_t2b => 'TEX CYRILLIC FONT ENCODING - T2B',
    fontools_t2c => 'TEX CYRILLIC FONT ENCODING - T2C',
    fontools_t3  => 'TEX IPA ENCODING',
    fontools_ts3 => 'TEX IPA SYMBOL ENCODING',
);

#-----------------------------------------------------------------------
# Generate a command line for otftotfm
#-----------------------------------------------------------------------
sub make_cmdline {
    my $item = shift;

    if ( !$ARGV{ligatures} ) {
        @{$item->{features}} = grep { $_ ne 'liga' } @{$item->{features}};
    }
    return join ' ', 'otftotfm',
                     ($ARGV{manual} ? '--pl' : '--automatic'),
                     "--encoding=$item->{encoding}[0]",
                     set_targetdirs($item->{font}),
                     '--no-updmap',
                     ($item->{font}{filename} =~ m/[.]ttf\z/xmsi
                        ? '--no-type1'
                        : ''),
                     ($SCHEME{$item->{encoding}[0]}
                        ? qq(--coding-scheme="$SCHEME{$item->{encoding}[0]}")
                        : ''),
                     (map { "--feature=$_" } @{$item->{features}}),
                     ($ARGV{ligatures}
                        ? ( '--ligkern="f i =: fi"',
                            '--ligkern="f l =: fl"',
                            '--ligkern="f f =: ff"',
                            '--ligkern="ff i =: ffi"',
                            '--ligkern="ff l =: ffl"' )
                        : ''),
                     $STYLE{$item->{style}}{extra},
                     $SHAPE{$item->{shape}}{extra},
                     $ARGV{extra},
                     qq("$item->{font}{filename}"),
                     $item->{fontname},
                     ;
}

#-----------------------------------------------------------------------
# Return a string with all "directory" options for otftotfm set
#-----------------------------------------------------------------------
sub set_targetdirs {
    my $font = shift;

    my $family = $font->{family};
    my ($fonttype) = map { lc $_ } $font->{filename} =~ m/[.]([ot]tf)\z/xmsi;

    my @FILETYPES = qw(tfm vf);
    push @FILETYPES, $fonttype eq 'otf' ? qw(type1)
                   : $fonttype eq 'ttf' ? qw(truetype)
                   :                      qw(type1 truetype)
                   ;

    my %dir = map { (
                        $_ => File::Spec->catdir(
                            $ARGV{target},
                            'fonts',
                            $_,
                            $ARGV{vendor},
                            $ARGV{typeface} || $family
                        )
                    )
                  }
                  @FILETYPES;

    $dir{$_}
        = File::Spec->catdir(
            $ARGV{target}, 'fonts', $_, 'dvips', $ARGV{typeface} || $family
        ) for qw(enc map);

    File::Path::make_path(values %dir) unless $ARGV{dryrun};

    my $result = join ' ', map { qq(--${_}-directory="$dir{$_}") } @FILETYPES;
    $result .= qq( --encoding-directory="$dir{enc}" --map-file=")
                . File::Spec->catfile($dir{map}, "${family}.map")
                . '"';

    return $result;
}


############################################################################


package ARGV;

my $USAGE =<<'END_USAGE';

'autoinst' is a wrapper around Eddie Kohler's TypeTools
(http://www.lcdf.org/type/), for installing OpenType fonts in LaTeX.

Usage: autoinst [options] font[s]

Possible options:
    -encoding=ENC[,ENC]*    Specify text encoding(s) (default: OT1,LY1,T1)

    -(no)lining             Toggle creation of lining digits
    -(no)oldstyle           Toggle creation of oldstyle digits
    -(no)proportional       Toggle creation of proportional digits
    -(no)tabular            Toggle creation of tabular digits
    -(no)ts1                Toggle creation of TS1 fonts
    -(no)smallcaps          Toggle creation of smallcaps shape
    -(no)swash              Toggle creation of swash shape
    -(no)titling            Toggle creation of titling shape
    -(no)superiors          Toggle creation of fonts with superior characters
    -noinferiors
    -inferiors=[none|auto|subs|sinf|dnom]
                            Use this style for subscripts (see docs)
    -(no)ornaments          Toggle creation of ornament fonts
    -(no)fractions          Toggle creation of fonts with digits for fractions
    -(no)ligatures          Toggle manual addition of f-ligatures to font

    -serif                  Install font as serif font
    -sanserif               Install font as sanserif font
    -typewriter             Install font as typewriter font

    -defaultlining
    -defaultoldstyle        Specify which figure style should be
    -defaulttabular             considered 'default' for this font; see docs
    -defaultproportional

    -extra="TEXT"           Add TEXT to the command for 'otftotfm'
    -target="DIRECTORY"     Install files into specified TEXMF tree
    -vendor="VENDOR"        Only used for naming directories
    -typeface="TYPEFACE"    Only used for naming directories
    -(no)updmap             Toggle running of updmap
    -manual                 Manual mode (see documentation)

    -(no)figurekern         Keep or remove kerns between tabular figures
    -(no)mergewidths        Merge condended/extended subfamilies with main
    -(no)mergeweights       Merge separate weights with main family
    -(no)mergeshapes        Merge separate shapes with main family

    -nfssweight=xx=yyyy     Map the "xx" NFSS code to the "yyyy" weight
    -nfsswidth=xx=yyyy      Map the "xx" NFSS code to the "yyyy" width

    -help                   Print this text and exit
    -doc                    Print the complete documentation and exit
    -dryrun                 Don't generate fonts, only log what would be done
    -logfile="FILE"         Write log to "FILE" (default: <fontfamily>.log)
    -verbose                Print more data to log file
                            (repeat for even higher verbosity)
    -version                Print version number and exit
    font[s]                 The fonts (.otf or .ttf format) to install.

Please report any bugs or suggestions to <marcpenninga@gmail.com>.
END_USAGE

# Default values for the command-line arguments
%ARGV = (
    encoding        => 'OT1,LY1,T1',
    textcomp        => '2',     # 0 = no, 1 = yes, 2 = ('T1' ? yes : no)
    lining          => '1',     # 0 = no, 1 = yes
    oldstyle        => '1',     # 0 = no, 1 = yes
    proportional    => '1',     # 0 = no, 1 = yes
    tabular         => '1',     # 0 = no, 1 = yes
    smallcaps       => '1',     # 0 = no, 1 = yes
    swash           => '1',     # 0 = no, 1 = yes
    titling         => '1',     # 0 = no, 1 = yes
    superiors       => '1',     # 0 = no, 1 = yes
    inferiors       => 'none',  # values: none, auto, subs, sinf, dnom
    ornaments       => '1',     # 0 = no, 1 = yes
    fractions       => '0',     # 0 = no, 1 = yes
    ligatures       => '2',     # 0 = no, 1 = yes, 2 = ('tt' ? no : yes)
    nfss            => '',
    fig_height      => 'lnum',
    fig_width       => 'tnum',
    extra           => '',
    target          => '',
    vendor          => 'lcdftools',
    typeface        => '',
    updmap          => '1',     # 0 = no, 1 = yes
    manual          => '0',     # 0 = no, 1 = yes
    dryrun          => '0',     # 0 = no, 1 = yes
    logfile         => '',
    figurekern      => '1',     # 0 = no, 1 = yes
    mergewidths     => '1',     # 0 = no, 1 = yes
    mergeweights    => '1',     # 0 = no, 1 = yes
    mergeshapes     => '1',     # 0 = no, 1 = yes
    verbose         => 0,
    nfsswidth       => [],
    nfssweight      => [],
);

#-----------------------------------------------------------------------
# Process command-line arguments
#-----------------------------------------------------------------------
sub parse_options {
    $ARGV{cmdline} = join ' ', ($0, @ARGV);

    Getopt::Long::GetOptions(
        'help|?'              =>  sub { print $USAGE; exit },
        'version'             =>  sub { print "$VERSION\n"; exit },
        'doc'                 =>  sub { Pod::Usage::pod2usage(-verbose => 2) },
        'encoding=s'          => \$ARGV{encoding},
        'ts1!'                => \$ARGV{textcomp},
        'lining!'             => \$ARGV{lining},
        'oldstyle!'           => \$ARGV{oldstyle},
        'proportional!'       => \$ARGV{proportional},
        'tabular!'            => \$ARGV{tabular},
        'smallcaps!'          => \$ARGV{smallcaps},
        'swash!'              => \$ARGV{swash},
        'titling!'            => \$ARGV{titling},
        'superiors!'          => \$ARGV{superiors},
        'noinferiors'         =>  sub { $ARGV{inferiors} = 'none' },
        'inferiors:s'         => \$ARGV{inferiors},
        'ornaments!'          => \$ARGV{ornaments},
        'fractions!'          => \$ARGV{fractions},
        'ligatures!'          => \$ARGV{ligatures},
        'serif'               =>  sub { $ARGV{nfss} = 'rm' },
        'sanserif'            =>  sub { $ARGV{nfss} = 'sf' },
        'typewriter'          =>  sub { $ARGV{nfss} = 'tt' },
        'defaultlining'       =>  sub { $ARGV{fig_height} = 'lnum' },
        'defaultoldstyle'     =>  sub { $ARGV{fig_height} = 'onum' },
        'defaulttabular'      =>  sub { $ARGV{fig_width}  = 'tnum' },
        'defaultproportional' =>  sub { $ARGV{fig_width}  = 'pnum' },
        'extra=s'             => \$ARGV{extra},
        'target=s'            => \$ARGV{target},
        'vendor=s'            => \$ARGV{vendor},
        'typeface=s'          => \$ARGV{typeface},
        'updmap!'             => \$ARGV{updmap},
        'dryrun'              => \$ARGV{dryrun},
        'manual'              => \$ARGV{manual},
        'figurekern!'         => \$ARGV{figurekern},
        'mergewidths!'        => \$ARGV{mergewidths},
        'mergeweights!'       => \$ARGV{mergeweights},
        'mergeshapes!'        => \$ARGV{mergeshapes},
        'mergesmallcaps!'     => \$ARGV{mergeshapes},
        'logfile=s'           => \$ARGV{logfile},
        'verbose+'            => \$ARGV{verbose},
        'nfssweight=s%'       => sub {
                                     my ( $ignored, $key, $values ) = @_;
                                     push @{$ARGV{nfssweight}},
                                          "$key=$values";
                                     my @values = split m/,/, lc $values;
                                     $key = q{} if $key eq 'm';
                                     $NFSS_WEIGHT{$key}
                                         = [ @values, @{$NFSS_WEIGHT{$key}} ];
                                     $ARGV{$key} = 'user-defined';
                                 },
        'nfsswidth=s%'        => sub {
                                     my ( $ignored, $key, $values ) = @_;
                                     my @values = split m/,/, lc $values;
                                     push @{$ARGV{nfsswidth}},
                                          "$key=$values";
                                     $key = q{} if $key eq 'm';
                                     $NFSS_WIDTH{$key}
                                         = [ @values, @{$NFSS_WIDTH{$key}} ];
                                     $ARGV{$key} = 'user-defined';
                                 },
    )
    or die "$USAGE";

    die "$USAGE" unless @ARGV;

    delete $SHAPE{smallcaps}     unless $ARGV{smallcaps};
    delete $SHAPE{swash}         unless $ARGV{swash};

    delete $STYLE{Titl}          unless $ARGV{titling};
    delete $STYLE{Sup}           unless $ARGV{superiors};
    delete $STYLE{Orn}           unless $ARGV{ornaments};
    delete @STYLE{qw(Numr Dnom)} unless $ARGV{fractions};
    $ARGV{inferiors} ||= 'auto';
    if    ($ARGV{inferiors} eq 'auto') { $STYLE{Inf}{reqd} = ['auto'] }
    elsif ($ARGV{inferiors} eq 'subs') { $STYLE{Inf}{reqd} = ['subs'] }
    elsif ($ARGV{inferiors} eq 'sinf') { $STYLE{Inf}{reqd} = ['sinf'] }
    elsif ($ARGV{inferiors} eq 'dnom') { $STYLE{Inf}{reqd} = ['dnom'] }
    else  {
        $ARGV{inferiors} = 'none';
        delete $STYLE{Inf}
    }

    $ARGV{encoding} =~ s/\s+//xmsg;
    my @textencodings = grep { $_ ne 'TS1' }
                             map { uc }
                                 split /,/, $ARGV{encoding};
    $ARGV{encoding} = \@textencodings;

    # TS1-encoded fonts are generated if:
    # - the user explicitly asked for TS1, or
    # - the text encodings contain T1 and the user didn't turn off TS1
    unless ( (grep { $_ eq 'T1' } @{$ARGV{encoding}} and $ARGV{textcomp})
             or $ARGV{textcomp} == 1 )
    {
        delete $SHAPE{textcomp};
    }

    # Determine NFSS classification of this family, if the user
    # didn't explicitly specify it
    eval {
        if ( !$ARGV{nfss} ) {
            # EXPERIMENTAL: parse "post" table to see if font is fixed-width
            my $cmd = qq(otfinfo --dump-table "post" "$ARGV[0]");
            open my $otfinfo, '-|:raw', $cmd
                or die "could not fork(): $!";
            my $post_table = do { local $/; <$otfinfo> };
            close $otfinfo
                or die "'$cmd' failed";

            my $is_fixed_pitch = unpack '@12N', $post_table;

            $ARGV{nfss} = $is_fixed_pitch                 ? 'tt'
                        : $ARGV[0] =~ m/mono(?!type)/xmsi ? 'tt'
                        : $ARGV[0] =~ m/sans/xmsi         ? 'sf'
                        :                                   'rm'
                        ;
        }
    } or warn "[WARNING]   $@";

    # If the user didn't explicitly say anything about ligatures,
    # activate them unless the font is a typewriter font.
    if ($ARGV{ligatures} == 2) {
        $ARGV{ligatures} = $ARGV{nfss} ne 'tt' ? 1 : 0;
    }
    # Fix the %STYLE table to take 'default' figure styles into account.
    $STYLE_DEFAULTS{$ARGV{fig_height}}();
    $STYLE_DEFAULTS{$ARGV{fig_width}}();

    delete @STYLE{qw(LF TLF)}    unless $ARGV{lining};
    delete @STYLE{qw(OsF TOsF)}  unless $ARGV{oldstyle};
    delete @STYLE{qw(LF OsF)}    unless $ARGV{proportional};
    delete @STYLE{qw(TLF TOsF)}  unless $ARGV{tabular};

    my $localtarget = File::Spec->catdir( Cwd->getcwd(), 'autoinst_output' );
    if ($ARGV{manual}) {
        warn "[WARNING]   Option '-target' overridden by '-manual'!\n"
            if $ARGV{target};
        $ARGV{target} = $localtarget;
        $ARGV{updmap} = 0;
    }
    elsif ($ARGV{target}) {
        $ARGV{updmap} = 0;
        warn <<"END_WARNING_TARGET_UPDMAP";
[WARNING]   The '-target' option may interfere with kpathsea and updmap;
            automatic calling of updmap has been disabled.
            Please call updmap manually.
END_WARNING_TARGET_UPDMAP
    }
    elsif (!$ARGV{target}) {
        my $is_windows_os = ( $^O =~ /^MSWin/i );
        my $kpsepath = $is_windows_os
            ? eval { qx( kpsewhich -expand-var=\$TEXMFLOCAL;\$TEXMFHOME ) }
            : eval { qx( kpsewhich -expand-var='\$TEXMFLOCAL:\$TEXMFHOME' ) }
            ;
        if (!$kpsepath) {
            warn <<"END_WARNING_KPSEWHICH";
[WARNING]   Call to "kpsewhich" failed.
            Maybe your TeX system doesn't use the kpathsea library?

            Consider using the "-target" command line option
            to specify a TEXMF tree where autoinst should install all files.
END_WARNING_KPSEWHICH
        }
        else {
            my $pathsep = $is_windows_os ? ';' : ':';
            for my $dir ( split m/$pathsep/xms, substr $kpsepath, 0, -1 ) {
                if (-w $dir) { $ARGV{target} = $dir; last }
                if (-e $dir) { next }
                my $par = File::Spec->catdir( $dir, File::Spec->updir() );
                if (-w $par) { $ARGV{target} = $dir; last }
            }
        }
    }

    if (!$ARGV{target}) {
        $ARGV{target} = $localtarget;
        $ARGV{updmap} = 0;
        warn <<"END_WARNING_DUMPING_FILES";
[WARNING]   No user-writable TEXMF-tree found!

            I'm putting all generated files in "$ARGV{target}".

            Please install these files into a suitable TEXMF directory,
            update the filename database and run 'updmap' (or similar);
            see your TeX installation's documentation.
END_WARNING_DUMPING_FILES
    }

    if ($ARGV{target} =~ m/[ ]/xms) {
        warn <<"END_WARNING_SPACES_IN_PATHS";
[WARNING]   The pathname of your target directory contains spaces:
                "$ARGV{target}"
            If you experience any problems, try re-running autoinst
            on a target directory without spaces in its name.
END_WARNING_SPACES_IN_PATHS
    }

    if (!$ARGV{figurekern}) {
        my @digits = qw(zero one two three four five six seven eight nine);
        my $tkern
            = join ' ', map { my $left = $_;
                              map { qq(--ligkern="$left {} $_") } @digits
                            }
                            @digits;

        $STYLE{TLF}{extra}  = $tkern;
        $STYLE{TOsF}{extra} = $tkern;
    }
}


############################################################################


package Fontinfo;

#-----------------------------------------------------------------------
# Return ref to hash with all relevant info about this font
#-----------------------------------------------------------------------
sub parse_fontinfo {
    my ($filename) = @_;

    my $info = {
        filename     => $filename,
        width        => 'regular',
        weight       => 'regular',
        shape        => 'roman',
        minsize      => 0,
        maxsize      => 0,
        is_smallcaps => 0,
        weight_class => 0,
        width_class  => 0,
    };

    parse_basicinfo($info);
    parse_sizeinfo($info);
    parse_features($info);

    return $info;
}

# table for converting digits in font names to text (safer for LaTeX)
my @DIGITS = qw(Zero One Two Three Four Five Six Seven Eight Nine);

#-----------------------------------------------------------------------
# Get some basic info (family, weight, width, shape) about this font
#-----------------------------------------------------------------------
sub parse_basicinfo {
    my $info = shift;

    my $cmd = qq(otfinfo --info "$info->{filename}");
    open my $otfinfo, '-|', $cmd
        or die "[ERROR]     Could not fork(): $!";
    my %data = map { my ($k,$v) = m/\A\s* ([^:]+?) \s*:\s* ([^\r\n]+)/xms;
                     $k =~ s/\s+//xmsg;
                     $v =~ s/\s+//xmsg;
                     (lc $k => $v);
                   }
                   grep { m/\A\s* [^:]+? \s*:\s* [^\r\n]+/xms } <$otfinfo>;
    close $otfinfo
        or die "[ERROR]     '$cmd' failed.";

    $data{family}    =  $data{preferredfamily} || $data{family};
    $data{subfamily} =  $data{preferredsubfamily} || $data{subfamily};
    $data{fullname}  =~ s/\A$data{family}//xms;
    $data{fullname}  =  lc $data{fullname};

    # clean up family name (it's used in LaTeX command names)
    $data{family}    =~ s/\A(?: Adobe | DTL | FF | ITC | LT | MT)//xms;
    $data{family}    =~ s/(?: LT | MT)(?: Std | Pro )\z//xms;
    $data{family}    =~ s/ Std \z//xms;
    $data{family}    =~ s/(\d)/$DIGITS[$1]/xmsge;
    $data{family}    =~ s/[^A-Za-z]+//xmsg;

    # Sometimes the relevant info is in Fullname, sometimes in Subfamily;
    # so we need to test against both
    my $fullinfo = lc "$data{subfamily} | $data{fullname}";

    # We need to be careful when parsing the font info; in particular
    # we must parse strings like 'UltraCondensed' as 'Regular' weight
    # and 'UltraCondensed' width, not as 'Ultra' weight and 'Condensed' width.
    # The following rules should prevent accidents:
    # 1.  Search for matching widths before matching weights
    #     (as none of the widths is a proper substring of some weight)
    # 2.  Remove any recognised search string from the 'fullinfo'
    # 3.  Test the weights 'medium' and 'regular' *last*, since these strings
    #     may also occur in Subfamily without indicating the weight;
    #     so we only take them to mean weight if we find no other hit.
    for my $width ( mqrs(@WIDTHS) ) {
        if ($fullinfo =~ m/${width}/xms) {
            $info->{width} = $width;
            my $widths = join '|', mqrs(@WIDTHS);
            $fullinfo =~ s/${widths}//gxmsi;
            last;
        }
    }
    for my $weight ( mqrs(@WEIGHTS), qw(medium regular) ) {
        if ($fullinfo =~ m/${weight}/xms) {
            $info->{weight} = $weight;
            my $weights = join '|', mqrs(@WEIGHTS), qw(medium regular);
            $fullinfo =~ s/${weights}//gxmsi;
            last;
        }
    }
    for my $shape ( mqrs( keys %NFSS_SHAPE ) ) {
        if ($fullinfo =~ m/${shape}/xms) {
            $info->{shape} = $shape;
            my $shapes = join '|', mqrs( keys %NFSS_SHAPE );
            $fullinfo =~ s/${shapes}//gxmsi;
            last;
        }
    }

    # In many font families, each font is in a subfamily of its own;
    # so we remove width, weight and shape from the 'subfamily' value.
    $data{subfamily} =~ s/$info->{width}//xmsi;
    $data{subfamily} =~ s/$info->{weight}//xmsi;
    $data{subfamily} =~ s/$info->{shape}//xmsi;

    $info->{name}      = $data{postscriptname};
    $info->{family}    = $data{family};
    $info->{subfamily} = $data{subfamily};

    # Take care to unabbreviate weight and width; CondensedUltra fonts
    # might end up as 'ultracondensed' instead of 'ultrablackcondensed'!
    if (exists $UNABBREVIATE{ $info->{width} }) {
        $info->{width} = $UNABBREVIATE{ $info->{width} };
    }
    if (exists $UNABBREVIATE{ $info->{weight} }) {
        $info->{weight} = $UNABBREVIATE{ $info->{weight} };
    }
    if (exists $UNABBREVIATE{ $info->{shape} }) {
        $info->{shape} = $UNABBREVIATE{ $info->{shape} };
    }

    if ($ARGV{mergeshapes}) {
        my $shapes = join '|', mqrs( qw(it italic) );
        if ( $info->{family} =~ m/(.+?) (${shapes}) \z/xmsi
                and ( $info->{shape} eq 'regular'
                   or $info->{shape} eq ( $UNABBREVIATE{ lc($2) } // lc($2) )
                )
        ) {
            $info->{family} = $1;
            $info->{shape}  = $UNABBREVIATE{ lc($2) } // lc($2);
        }

        $shapes = join '|', mqrs( qw(sc smallcaps) );
        if ( $info->{family} =~ m/(.+?) ( -? ${shapes}) \z/xmsi ) {
            $info->{family}       = $1;
            $info->{is_smallcaps} = 1;
        }
        if ( $info->{name} =~ m/(.+?) ( -? ${shapes}) \z/xmsi ) {
            $info->{name}         = $1;
            $info->{is_smallcaps} = 1;
        }
    }

    # Some font families put different widths into separate families;
    # we provide an option to merge these with the 'main' font family.
    if ($ARGV{mergewidths}) {
        my $widths = join '|', mqrs(@WIDTHS);
        if ( $info->{family} =~ m/(.+?) (${widths}) \z/xmsi
                and ( $info->{width} eq 'regular'
                   or $info->{width} eq ( $UNABBREVIATE{ lc($2) } // lc($2) )
                )
        ) {
            $info->{family} = $1;
            $info->{width}  = $UNABBREVIATE{ lc($2) } // lc($2);
        }
    }

    # Some font families put extreme weights into separate families;
    # we provide an option to merge these with the 'main' font family.
    # Note that if the font's family name includes the word "Text"
    # (as in Libre Caslon Text) this should not be mistaken for a weight.
    if ($ARGV{mergeweights}) {
        my $weights = join '|', mqrs(@WEIGHTS), qw(medium regular);
        if ( $info->{family} =~ m/text \z/xmsi ) {
            $weights =~ s/text[|]?//xms;
        }
        if ( $info->{family} =~ m/(.+?) (${weights}) \z/xmsi
                and ( $info->{weight} eq 'regular'
                   or $info->{weight} eq ( $UNABBREVIATE{ lc($2) } // lc($2) )
                )
        ) {
            $info->{family} = $1;
            $info->{weight} = $UNABBREVIATE{ lc($2) } // lc($2);
        }
    }

    # Strip off the "Text" from family names that contain this string.
    # This was a crude way to fix a bug in the previous paragraph;
    # it's unnecessary now, but we don't want to break the old behaviour.
    $info->{family} =~ s/text \z//xmsi;

    $info->{basicshape} = $NFSS_SHAPE{$info->{shape}};

    # We define 'series' as 'weight + width'. This matches NFSS,
    # but contradicts how most fonts are named (which is 'width + weight').
    $info->{series}
        = ($info->{width}  eq 'regular') ? $info->{weight}
        : ($info->{weight} eq 'regular') ? $info->{width}
        :                                  $info->{weight} . $info->{width}
        ;

    # EXPERIMENTAL: we extract the usWeightClass and usWidthClass
    # properties from the font's OS/2 table
    eval {
        my $cmd = qq(otfinfo --dump-table "OS/2" "$info->{filename}");
        open $otfinfo, '-|:raw', $cmd
            or die "could not fork(): $!";
        my $os2_table = do { local $/; <$otfinfo> };
        close $otfinfo
            or die "'$cmd' failed";
        @{$info}{qw(weight_class width_class)} = unpack '@4n @6n', $os2_table;
    } or warn "[WARNING]   $@";

    return;
}

#-----------------------------------------------------------------------
# Re-order argument list so that longer strings come before shorter ones
#
# This is needed when matching a string against a list of patterns
# where some of the patterns may be substrings of other patterns;
# then we want the *longest* matching pattern, so we test the patterns
# in descending order of length.
#-----------------------------------------------------------------------
sub mqrs {
    return map { quotemeta } reverse sort { length($a) <=> length($b) } @_;
}

#-----------------------------------------------------------------------
# Fill the 'feature' field in the 'info' struct with a list of
# all features this font supports
#-----------------------------------------------------------------------
sub parse_features {
    my $info = shift;

    my $cmd = qq(otfinfo --features "$info->{filename}");
    open my $otfinfo, '-|', $cmd
        or die "[ERROR]     Could not fork(): $!";
    %{$info->{feature}} = map { (substr($_, 0, 4) => 1) } <$otfinfo>;
    close $otfinfo
        or die "[ERROR]     '$cmd' failed.";

    $cmd = qq(otfinfo --tables "$info->{filename}");
    open $otfinfo, '-|', $cmd
        or die "[ERROR]     Could not fork(): $!";
    $info->{feature}{kern} = 1 if grep { m/\d+ \s+ kern/xms } <$otfinfo>;
    close $otfinfo
        or die "[ERROR]     '$cmd' failed.";

    return;
}

#-----------------------------------------------------------------------
# Fill the 'minsize' and 'maxsize' fields in the 'info' struct
# with limits of optical design size range of this font
#-----------------------------------------------------------------------
sub parse_sizeinfo {
    my $info = shift;

    my $cmd = qq(otfinfo --optical-size "$info->{filename}");
    open my $otfinfo, '-|', $cmd
        or die "[ERROR]     Could not fork(): $!";

    if (my ($minsize, $maxsize)
        = <$otfinfo> =~ m/[(] ([\d.]+) \s* pt, \s*
                              ([\d.]+) \s* pt  \s* []]/xms
    ) {
    # fix some known bugs
        if ($info->{name} eq 'GaramondPremrPro-It'
            && $minsize == 6 && $maxsize == 8.9)
        {
            ($minsize, $maxsize) = (8.9, 14.9);
        }
        elsif ($info->{family} eq 'KeplerStd'
            && $info->{subfamily} =~ m/Caption/xms
            && $minsize == 8.9 && $maxsize == 13.9)
        {
            ($minsize, $maxsize) = (6, 8.9);
        }
        elsif ($info->{family} eq 'KeplerStd'
            && $info->{subfamily} =~ m/Subhead/xms
            && $minsize == 8.9 && $maxsize == 13.9)
        {
            ($minsize, $maxsize) = (13.9, 23);
        }
        elsif ($info->{family} eq 'KeplerStd'
            && $info->{subfamily} =~ m/Display/xms
            && $minsize == 8.9 && $maxsize == 13.9)
        {
            ($minsize, $maxsize) = (23, 72);
        }

        @{$info}{qw(minsize maxsize)} = ($minsize, $maxsize);
    }
    close $otfinfo
        or die "[ERROR]     '$cmd' failed.";

    return;
}


# Error messages, used in assert_unique()
my $ERR_DETAIL =<<'END_ERR_DETAIL';
[ERROR]     I've parsed both %s
                         and %s as

            Family:     %s
            Weight:     %s
            Width:      %s
            Shape:      %s
            Size:       %s-%s

END_ERR_DETAIL

my $ERR_PARSE =<<'END_ERR_PARSE';
[ERROR]     I failed to parse all fonts in a unique way;
            presumably some fonts have unusual widths, weights or shapes.

            Try one of the following:
            -   Run 'autoinst' on a smaller set of fonts,
                omitting the ones that weren't parsed correctly;
            -   Add the missing widths, weights or shapes to the tables
                'NFSS_WIDTH', 'NFSS_WEIGHT' or 'NFSS_SHAPE' near the top
                of the source code;

            Please also send a bug report to the author.
END_ERR_PARSE

#-----------------------------------------------------------------------
# Assert all parsed font infos are unique
#-----------------------------------------------------------------------
sub assert_unique {
    my @fonts = @_;

    # These attributes should uniquely identify each font
    my @attributes
        = qw(family weight width shape minsize maxsize is_smallcaps);

    my (%seen, $err_details);
    for my $font (@fonts) {
        my $key = join "\x00", @{$font}{ @attributes };

        if ($seen{$key}) {
            $err_details .= sprintf $ERR_DETAIL,
                                    $seen{$key}{filename},
                                    $font->{filename},
                                    @{$font}{ @attributes };
        }
        else {
             $seen{$key} = $font;
        }
    }

    # Die with detailed error message if the font infos aren't unique
    if ($err_details) {
        print {$ARGV{logfile}} $err_details, $ERR_PARSE;
        die $err_details, $ERR_PARSE;
    }

    my %fontfamilies = map { ( $_->{family} => 1 ) } @fonts;
    my @fontfamilies = keys %fontfamilies;
    if (scalar @fontfamilies > 1) {
        warn <<"END_WARNING_MULTIPLE_FAMILIES";
[WARNING]   Your fonts seem to belong to multiple families:

                @{[ join( "\n" . q{ } x 16, sort @fontfamilies ) ]}

            "autoinst" performs best when installing one family at a time.
            If you think your fonts should all be in the same family,
            you may have found a bug in autoinst's font info parsing.

            Please see the log file for details on how the fonts' families,
            weights, widths and shapes were parsed.
END_WARNING_MULTIPLE_FAMILIES

        # increase verbosity if necessary, to log all relevant info
        if ($ARGV{verbose} < 1) {
            $ARGV{verbose} = 1;
        }
    }

    return 1;
}

#-----------------------------------------------------------------------
# Generate (font-specific) encoding vector for ornament glyphs
#-----------------------------------------------------------------------
sub get_orn {
    my ($font) = @_;

    my $fontfile = $font->{filename};
    my $enc_name = $font->{family} . '_orn';

    if (not -e "$enc_name.enc") {
        # Default ornament names: 'orn.' plus three digits
        my @encoding = map { sprintf "orn.%03d", $_ } 1 .. 256;

        my $cmd = qq(otfinfo --glyphs "$font->{filename}");
        open my $otfinfo, '-|', $cmd
            or die "[ERROR]     Could not fork(): $!";
        chop(my @glyphnames = <$otfinfo>);
        close $otfinfo
            or die "[ERROR]     '$cmd' failed.";

        # Test for some known alternative names (probably not exhaustive)
        my @ornaments
            = sort grep { m/\A (?: orn|u2022[.]|word[.]|hand|bullet[.])
                            | [.]orn \z/xms
                        }
                        @glyphnames;

        @encoding[0 .. $#ornaments] = @ornaments;

        open my $ORN, '>', "$enc_name.enc"
            or die "[ERROR]     Can't create '$enc_name.enc': $!";

        print {$ORN} "/$font->{family}OrnamentEncoding [\n";
        map { print {$ORN} "    /$_\n" } @encoding[0 .. 255];
        print {$ORN} "] def\n";
        close $ORN;
    }

    return $enc_name;
}

#-----------------------------------------------------------------------
# Handle the -inferiors=auto option
#-----------------------------------------------------------------------
sub handle_auto_inferiors {
    FONT:
    for my $font (@_) {
        for my $inf (qw(subs sinf dnom)) {
            if (exists $font->{feature}{$inf}) {
                $ARGV{inferiors} = $inf;
                $STYLE{Inf}{reqd} = [$inf];
                last FONT;
            }
        }
    }

    # If we didn't find any inferior figures, delete the 'Inf' entry
    # from the %STYLE table to indicate we don't want to generate this style.
    delete $STYLE{Inf} if $ARGV{inferiors} eq 'auto';
}


############################################################################


package LaTeX;

#-----------------------------------------------------------------------
# Create a LaTeX style file
#-----------------------------------------------------------------------
sub write_stylefile {
    my ($fam, $data) = @_;

    my %seen = %{ get_keys($data) };

    my $fn = sprintf "%s.sty", $fam;
    my $dir = File::Spec->catdir(
        $ARGV{target}, 'tex', 'latex', $ARGV{typeface} || $fam);
    File::Path::make_path($dir);
    $fn = File::Spec->catfile($dir, $fn);
    open my $STY, '>', $fn
        or die "[ERROR]     Can't create '$fn': $!";
    binmode $STY;

    print {$STY} <<"END_STY_HEADER";
%% Generated by autoinst on $TODAY
%%
\\NeedsTeXFormat{LaTeX2e}
\\ProvidesPackage{$fam}
    [$TODAY (autoinst)  Style file for $fam.]

END_STY_HEADER

    my $enc = join ',', grep { $_ ne 'OT1' } @{$ARGV{encoding}};

    print {$STY} "\\RequirePackage[$enc]{fontenc}\n" if $enc;
    print {$STY} "\\RequirePackage{textcomp}\n" if $seen{TS1};

    print {$STY} <<'END_STY_FONTAXES_START';
\IfFileExists{mweights.sty}{\RequirePackage{mweights}}{}
\IfFileExists{fontaxes.sty}{
    \RequirePackage{fontaxes}
END_STY_FONTAXES_START


    if ($seen{nw} or $seen{sw}) {
        print {$STY} <<'END_STY_FONTAXES_SW';
    \DeclareRobustCommand\swshape{\not@math@alphabet\swshape\relax
        \fontprimaryshape\itdefault\fontsecondaryshape\swdefault\selectfont}
    \fa@naming@exception{shape}{{n}{sw}}{nw}
    \fa@naming@exception{shape}{{it}{sw}}{sw}

END_STY_FONTAXES_SW
    }

    if ($seen{Sup}) {
        print {$STY} <<'END_STY_FONTAXES_SUP';
    \fa@naming@exception{figures}{{superior}{proportional}}{Sup}
    \fa@naming@exception{figures}{{superior}{tabular}}{Sup}
    \def\sufigures{\@nomath\sufigures
        \fontfigurestyle{superior}\selectfont}
    \DeclareTextFontCommand{\textsu}{\sufigures}
    \let\textsuperior\textsu

END_STY_FONTAXES_SUP
    }

    if ($seen{Inf}) {
        print {$STY} <<'END_STY_FONTAXES_INF';
    \fa@naming@exception{figures}{{inferior}{proportional}}{Inf}
    \fa@naming@exception{figures}{{inferior}{tabular}}{Inf}
    \def\infigures{\@nomath\infigures
        \fontfigurestyle{inferior}\selectfont}
    \DeclareTextFontCommand{\textin}{\infigures}
    \let\textinferior\textin

END_STY_FONTAXES_INF
    }

    if ($seen{Titl}) {
        print {$STY} <<'END_STY_FONTAXES_TITL';
    \fa@naming@exception{figures}{{titlingshape}{proportional}}{Titl}
    \fa@naming@exception{figures}{{titlingshape}{tabular}}{Titl}
    \def\tlshape{\@nomath\tlshape
        \fontfigurestyle{titlingshape}\selectfont}
    \DeclareTextFontCommand{\texttl}{\tlshape}
    \let\texttitling\texttl

END_STY_FONTAXES_TITL
    }

    if ($seen{Orn}) {
        print {$STY} <<'END_STY_FONTAXES_ORN';
    \fa@naming@exception{figures}{{ornament}{proportional}}{Orn}
    \fa@naming@exception{figures}{{ornament}{tabular}}{Orn}
    \def\ornaments{\@nomath\ornaments
        \fontencoding{U}\fontfigurestyle{ornament}\selectfont}
    \DeclareTextFontCommand{\textornaments}{\ornaments}
    \providecommand{\ornament}[1]{\textornaments{\char##1}}

END_STY_FONTAXES_ORN
    }

    if ($seen{Numr}) {
        print {$STY} <<'END_STY_FONTAXES_NUMR';
    \fa@naming@exception{figures}{{numerators}{proportional}}{Numr}
    \fa@naming@exception{figures}{{numerators}{tabular}}{Numr}

END_STY_FONTAXES_NUMR
    }

    if ($seen{Dnom}) {
        print {$STY} <<'END_STY_FONTAXES_DNOM';
    \fa@naming@exception{figures}{{denominators}{proportional}}{Dnom}
    \fa@naming@exception{figures}{{denominators}{tabular}}{Dnom}

END_STY_FONTAXES_DNOM
    }

    print {$STY} "}{}\n\n";

    print {$STY} <<"END_STY_XKEYVAL";
\\IfFileExists{xkeyval.sty}{
    \\newcommand*{\\$fam\@scale}{1}
    \\RequirePackage{xkeyval}
    \\DeclareOptionX{scale}{\\renewcommand*{\\$fam\@scale}{##1}}
    \\DeclareOptionX{scaled}{\\renewcommand*{\\$fam\@scale}{##1}}
}{
    \\let\\DeclareOptionX\\DeclareOption
    \\let\\ExecuteOptionsX\\ExecuteOptions
    \\let\\ProcessOptionsX\\ProcessOptions
}

END_STY_XKEYVAL

    if ($seen{LF} or $seen{TLF}) {
        print {$STY}
            "\\DeclareOptionX{lining}{\\edef\\$fam\@figurestyle{LF}}\n";
    }
    if ($seen{OsF} or $seen{TOsF}) {
        print {$STY}
            "\\DeclareOptionX{oldstyle}{\\edef\\$fam\@figurestyle{OsF}}\n";
    }
    if ($seen{TLF} or $seen{TOsF}) {
        print {$STY}
            "\\DeclareOptionX{tabular}{\\edef\\$fam\@figurealign{T}}\n";
    }
    if ($seen{LF} or $seen{OsF}) {
        print {$STY}
            "\\DeclareOptionX{proportional}{\\edef\\$fam\@figurealign{}}\n";
    }

    print {$STY} <<"END_STY_MAINFONT";
\\DeclareOptionX{mainfont}{
    \\renewcommand{\\familydefault}{\\$ARGV{nfss}default}
}
END_STY_MAINFONT

    my $defaults
        = $seen{OsF}  ? 'oldstyle,proportional'
        : $seen{TOsF} ? 'oldstyle,tabular'
        : $seen{LF}   ? 'lining,proportional'
        : $seen{TLF}  ? 'lining,tabular'
        :               die "[ERROR]     Internal bug, please report!";

    my $default_bold;
    for my $series (qw(heavy black extrabold demibold semibold bold)) {
        if ( $seen{$series} ) {
            print {$STY}
                "\\DeclareOptionX{$series}{\\edef\\bfseries\@$ARGV{nfss}",
                "{$series}}\n";
            $default_bold = $series;
        }
    }
    $defaults .= ",$default_bold" if $default_bold;

    my $default_regular;
    for my $series (qw(medium book text regular)) {
        if ( $seen{$series} ) {
            print {$STY}
                "\\DeclareOptionX{$series}{\\edef\\mdseries\@$ARGV{nfss}",
                "{$series}}\n";
            $default_regular = $series;
        }
    }
    $defaults .= ",$default_regular" if $default_regular;

    print {$STY} <<"END_STYLE_REST";
\\ExecuteOptionsX{$defaults}
\\ProcessOptionsX\\relax

\\renewcommand*
    {\\$ARGV{nfss}default}
    {$fam-\\$fam\@figurealign\\$fam\@figurestyle}

\\endinput
END_STYLE_REST

    close $STY;

    return;
}

#-----------------------------------------------------------------------
# Walk a nested dictionary, return lookup table with all keys
#-----------------------------------------------------------------------
sub get_keys {
    my $dict = shift;
    my $seen = shift || {};

    while (my ($k, $v) = each %$dict) {
        $seen->{$k} = 1;
        get_keys($v, $seen) if ref $v eq 'HASH';
    }

    return $seen;
}

#-----------------------------------------------------------------------
# Create a .fd file for NFSS
#-----------------------------------------------------------------------
sub write_fdfile {
    my ($fam, $enc, $sty, $data) = @_;

    my $fn = sprintf "%s%s-%s.fd", $enc, $fam, $sty;
    my $dir = File::Spec->catdir(
        $ARGV{target}, 'tex', 'latex', $ARGV{typeface} || $fam);
    File::Path::make_path($dir);
    $fn = File::Spec->catfile($dir, $fn);
    open my $FD, '>', $fn
        or die "[ERROR]     Can't create '$fn': $!";
    binmode $FD;

    print {$FD} <<"END_FD_HEADER";
%% Generated by autoinst on $TODAY
%%
\\ProvidesFile{${enc}${fam}-${sty}.fd}
    [$TODAY (autoinst)  Font definitions for ${enc}/${fam}-${sty}.]

\\ifcsname s\@fct\@alias\\endcsname\\else
\\gdef\\s\@fct\@alias{\\sub\@sfcnt\\\@font\@aliasinfo}
\\gdef\\\@font\@aliasinfo#1{%
    \\\@font\@info{Font\\space shape\\space `\\curr\@fontshape'\\space will
        \\space be\\space aliased\\MessageBreak to\\space `\\mandatory\@arg'}%
}
\\fi

\\expandafter\\ifx\\csname ${fam}\@scale\\endcsname\\relax
    \\let\\${fam}\@\@scale\\\@empty
\\else
    \\edef\\${fam}\@\@scale{s*[\\csname ${fam}\@scale\\endcsname]}%
\\fi

\\DeclareFontFamily{${enc}}{${fam}-${sty}}{@{[
    $ARGV{nfss} eq 'tt' ? '\hyphenchar\font=-1' : ""
]}}

END_FD_HEADER

    while (my ($series, $fdseries) = each %$data) {
        print {$FD} "\n%   ----  $series  ----\n\n";
        while (my ($shape, $fdshape) = each %$fdseries) {
            print {$FD}
                "\\DeclareFontShape{$enc}{${fam}-${sty}}{$series}{$shape}{\n";
            my @sizes = sort { $a->[0] <=> $b->[0] }
                             @{$fdshape};
            $sizes[0][0] = $sizes[-1][1] = '';
            $sizes[$_][0] = $sizes[$_ - 1][1] for (1 .. $#sizes);
            for my $size (@sizes) {
                print {$FD} "      <$size->[0]-$size->[1]> ",
                            "\\${fam}\@\@scale $size->[2]\n";
            }
            print {$FD} "}{}\n\n";
        }

        # ssub italic for missing slanted, or vice versa
        while (my ($shape, $replace) = each %SSUB_SHAPE) {
            if (!exists $fdseries->{$shape} && exists $fdseries->{$replace}) {
                print {$FD} <<"END_SSUB_SHAPE";
\\DeclareFontShape{$enc}{${fam}-${sty}}{$series}{$shape}{
      <-> ssub * ${fam}-${sty}/${series}/${replace}
}{}

END_SSUB_SHAPE
                $fdseries->{$shape} = 1;
            }
        }
    }

    print {$FD} <<"END_COMMENT";
%
%  Extra 'alias' rules to map the standard NFSS codes to our fancy names
%
END_COMMENT
    my %seen;
    NFSSWEIGHT:
    for my $nfssweight (@NFSS_WEIGHT) {
        NFSSWIDTH:
        for my $nfsswidth (@NFSS_WIDTH) {
            my $nfssseries = ( $nfssweight . $nfsswidth) || 'm';

            for my $weight ( @{$NFSS_WEIGHT{$nfssweight}} ) {
                $weight = '' if $weight eq 'regular';
                for my $width ( @{$NFSS_WIDTH{$nfsswidth}} ) {
                    $width = '' if $width eq 'regular';
                    my $series = ( $weight . $width ) || 'regular';
                    if ( exists $data->{$series} ) {
                        print {$FD} "\n%   $nfssseries --> $series\n\n";
                        for my $shape ( keys %{$data->{$series}} ) {
                            print {$FD} <<"END_SSUB_SERIES";
\\DeclareFontShape{$enc}{${fam}-${sty}}{$nfssseries}{$shape}{
      <-> alias * ${fam}-${sty}/${series}/${shape}
}{}

END_SSUB_SERIES
                            $seen{$nfssseries}{$shape} = 1;
                        }
                        next NFSSWIDTH;
                    }
                }
            }
        }
    }

    # Add ssub rules to map bx to b
    for my $shape ( keys %{$seen{b}} ) {
        if ( !exists $seen{bx}{$shape} ) {
            print {$FD} <<"END_SSUB_BX";
\\DeclareFontShape{$enc}{${fam}-${sty}}{bx}{$shape}{
      <-> ssub * ${fam}-${sty}/b/${shape}
}{}

END_SSUB_BX
        }
    }

    print {$FD} "\\endinput\n";
    close $FD;

    return;
}


############################################################################


package main;

main();


__END__


############################################################################


    To create the documentation:

    pod2man --center="Marc Penninga" --release="fontools" --section=1 \
        autoinst - | groff -Tps -man - | ps2pdf - autoinst.pdf


=pod

=head1 NAME

autoinst - wrapper around the F<LCDF TypeTools>,
for installing and using OpenType fonts in LaTeX.


=head1 SYNOPSIS

B<autoinst> [I<options>] B<fontfile(s)>


=head1 DESCRIPTION

Eddie Kohler's I<LCDF TypeTools> are superb tools for installing
OpenType fonts in LaTeX, but they can be hard to use:
they need many, often long, command lines
and don't generate the F<fd> and F<sty> files LaTeX needs.
B<autoinst> simplifies the use of the I<TypeTools> for font installation
by generating and executing all commands for I<otftotfm>
and by creating and installing all necessary F<fd> and F<sty> files.

Given a family of font files (in F<otf> or F<ttf> format),
B<autoinst> will create several LaTeX font families:

=over 2

=over 3

=item -

Four text families (with lining and oldstyle digits,
each in both tabular and proportional variants),
all with the following shapes:

=over 2

=over 8

=item I<n>

Roman (i.e., upright) text

=item I<it>, I<sl>

Italic and slanted (sometimes called oblique) text

=item I<sc>

Small caps

=item I<scit>, I<scsl>

Italic and slanted small caps

=item I<sw>

Swash

=item I<nw>

"Upright swash"

=back

=back

=item -

For each T1-encoded text family:
a family of TS1-encoded symbol fonts, in roman, italic and slanted shapes.

=item -

Families with superiors, inferiors, numerators and denominators,
in roman, italic and slanted shapes.

=item -

Families with "Titling" characters;
these "... replace the default glyphs
with corresponding forms designed specifically for titling.
These may be all-capital and/or larger on the body,
and adjusted for viewing at larger sizes"
(according to the OpenType Specification).

=item -

An ornament family, also in roman, italic and slanted shapes.

=back

=back

Of course, if your fonts don't contain italics, oldstyle digits, small caps
etc., the corresponding shapes and families are not created.
In addition, the creation of most families and shapes can be controlled
by the user (see L</"COMMAND-LINE OPTIONS"> below).

These families use the I<FontPro> project's naming scheme:
I<< <FontFamily>-<Suffix> >>, where I<< <Suffix> >> is:

=over 8

=item I<LF>

proportional (i.e., figures have varying widths) lining figures

=item I<TLF>

tabular (i.e., all figures have the same width) lining figures

=item I<OsF>

proportional oldstyle figures

=item I<TOsF>

tabular oldstyle figures

=item I<Sup>

superior characters (note that most fonts have only an incomplete set of
superior characters: digits, some punctuation and the letters I<abdeilmnorst>;
normal forms are used for other characters)

=item I<Inf>

inferior characters; usually only digits and some punctuation,
normal forms for other characters

=item I<Titl>

Titling characters; see above.

=item I<Orn>

ornaments

=item I<Numr>

numerators

=item I<Dnom>

denominators

=back

The individual fonts are named I<< <FontName>-<suffix>-<shape>-<enc> >>,
where I<< <suffix> >> is the same as above (but in lowercase),
I<< <shape> >> is either empty, "sc" or "swash",
and I<< <enc> >> is the encoding (also in lowercase).
A typical name in this scheme would be "FiraSans-Light-osf-sc-ly1".


=head2 About the log file

B<autoinst> writes some info about what it thinks it's doing to a log file.
By default this is called F<< <fontfamily>.log >>,
but this choice can be overridden by the user;
see the I<-logfile> command-line option in L</"COMMAND-LINE OPTIONS"> below.
If this log file already exists, B<autoinst> will append its data to the end
rather than overwrite it.
Use the I<-verbose> command-line option to ask for more detailed info.


=head2 A note for MiKTeX users

Automatically installing the fonts into a suitable TEXMF tree
(as B<autoinst> tries to do by default) only works for TeX-installations
that use the F<kpathsea> library; with TeX distributions that implement
their own directory searching (such as MiKTeX), B<autoinst> will complain
that it cannot find the F<kpsewhich> program and move all generated files
into a subdirectory C<./autoinst_output/> of the current directory.
If you use such a TeX distribution, you should either move these files
to their correct destinations by hand, or use the I<-target> option
(see L</"COMMAND-LINE OPTIONS"> below) to manually specify a TEXMF tree.

Also, some OpenType fonts contain so many kerning pairs that the resulting
F<pl> and F<vpl> files are too big for MiKTeX's F<pltotf> and F<vptovf>;
the versions that come with W32TeX (F<http://www.w32tex.org>)
and TeXLive (F<http://tug.org/texlive>) don't seem to have this problem.


=head2 A note for MacTeX users

By default, B<autoinst> will try to install all generated files into
the $TEXMFLOCAL tree; when this directory isn't user-writable,
it will use the $TEXMFHOME tree instead.  Unfortunately, MacTeX's version
of C<updmap-sys> (which is called behind the scenes) doesn't search
in $TEXMFHOME, and hence MacTeX will not find the new fonts.

To remedy this, either run B<autoinst> as root (so that it can install
everything into $TEXMFLOCAL) or manually run C<updmap -user> to tell
TeX about the files in $TEXMFHOME.
The latter option does, however, have some caveats;
see F<https://tug.org/texlive/scripts-sys-user.html>.


=head2 Using the fonts in your LaTeX documents

B<autoinst> generates a style file for using the fonts in LaTeX documents,
named F<< <FontFamily>.sty >>. This style file also takes care of loading
the F<fontenc> and F<textcomp> packages.
To use the fonts, add the command C<<< \usepackage{I<< <FontFamily> >>} >>>
to the preamble of your document.

This style file defines a number of options:

=over 4

=item C<mainfont>

Redefine C<\familydefault> to make this font the main font
for the document.
This is a no-op if the font is installed as a serif font;
but if the font is installed as a sanserif or typewriter font,
this option saves you from having to redefine C<\familydefault>
yourself.

=item C<lining>, C<oldstyle>, C<tabular>, C<proportional>

Choose which figure style to use.
The defaults are "oldstyle" and "proportional" (if available).

=item C<<< scale=I<< <number> >> >>>

Scale the font by a factor of I<< <number> >>.
E.g., to increase the size of the font by 5%, use
C<<< \usepackage[scale=1.05]{I<< <FontFamily> >>} >>>.
May also be spelled C<scaled>.

This option is only available when you have the F<xkeyval> package installed.

=item C<medium>, C<book>, C<text>, C<regular>

Select the weight that LaTeX will use as the "regular" weight;
the default is C<regular>.

=item C<heavy>, C<black>, C<extrabold>, C<demibold>, C<semibold>, C<bold>

Select the weight that LaTeX will use as the "bold" weight;
the default is C<bold>.

=back

The previous two groups of options will only work if
you have the F<mweights> package installed.

The style file will also try to load the F<fontaxes> package
(on CTAN), which gives easy access to various font shapes and styles.
Using the machinery set up by F<fontaxes>, the generated style file
defines a number of commands (which take the text to be typeset as argument)
and declarations (which don't take arguments, but affect all text up to
the end of the current group) to access titling, superior and inferior
characters:


    DECLARATION     COMMAND         SHORT FORM OF COMMAND

    \tlshape        \texttitling    \texttl
    \sufigures      \textsuperior   \textsu
    \infigures      \textinferior   \textin


In addition, the C<\swshape> and C<\textsw> commands are redefined to place
swash on F<fontaxes>' secondary shape axis (F<fontaxes> places it on the
primary shape axis) to make them behave properly when nested, so that
C<\swshape\upshape> will give upright swash.

There are no commands for accessing the numerator and denominator
fonts; these can be selected using F<fontaxes>' standard commands,
e.g., C<\fontfigurestyle{numerator}\selectfont>.

The style file also provides a command C<<< \ornament{I<< <number> >>} >>>,
where C<<< I<< <number> >> >>> is a number from 0 to the total number of
ornaments minus one. Ornaments are always typeset using the current family,
series and shape. A list of all ornaments in a font can be created by
running LaTeX on the file F<nfssfont.tex> (part of a standard
LaTeX installation) and supplying the name of the ornament font.

To access ornament glyphs, B<autoinst> creates a font-specific encoding file
F<< <FontFamily>_orn.enc >>,
but only if that file doesn't yet exist in the current directory.
This is a deliberate feature that allows you to provide your own encoding
vector, e.g. if your fonts use non-standard glyph names for ornaments.

These commands are only generated for existing shapes and number styles;
no commands are generated for shapes and styles that don't exist,
or whose generation was turned off by the user.
Also these commands are built on top of F<fontaxes>, so if that package
cannot be found, you're limited to using the lower-level commands from
standard NFSS (C<\fontfamily>, C<\fontseries>, C<\fontshape> etc.).

By default, B<autoinst> generates text fonts with OT1, LY1 and T1
encodings, and the generated style files use T1 as the default text encoding.
Other encodings can be chosen using the I<-encoding> option
(see L</"COMMAND-LINE OPTIONS"> below).


=head2 NFSS codes

LaTeX's New Font Selection System (NFSS)
identifies fonts by a combination of family,
series (the concatenation of weight and width), shape and size.
B<autoinst> parses the font's metadata
(more precisely: the output of C<otfinfo --info>)
to determine these parameters.
When this fails (usually because the font family contains uncommon weights,
widths or shapes),
B<autoinst> ends up with different fonts having the I<same> values
for these font parameters; such fonts cannot be used in NFSS,
since there's no way distinguish them.
When B<autoinst> detects such a situation, it will print an error message
and abort.
If that happens, either rerun B<autoinst> on a smaller set of fonts,
or add the missing widths, weights and shapes to the tables C<NFSS_WIDTH>,
C<NFSS_WEIGHT> and C<NFSS_SHAPE>, near the top of the source code.
Please also send a bug report (see L<AUTHOR> below).

The mapping of shapes to NFSS codes is done using the following table:

    SHAPE                               CODE
    --------------------------------    ----
    Roman, Upright                      n
    Italic                              it
    Oblique, Slant(ed), Incline(d)      sl

(I<Exception:> Adobe Silentium Pro contains two Roman shapes;
we map the first of these to "n", for the second one we (ab)use the "it" code
as this family doesn't contain an Italic shape.)

The mapping of weights and widths to NFSS codes is a more complex,
two-step proces.
In the first step, all fonts are assigned a "series" name that is simply
the concatenation of its weight and width
(after expanding any abbreviations and converting to lowercase).
A font with "Cond" width and "Ultra" weight will then be known
as "ultrablackcondensed".

In the second step, B<autoinst> tries to map all combinations of NFSS codes
(ul, el, l, sl, m, sb, b, eb and ub for weights;
uc, ec, c, sc, m, sx, x, ex and ux for widths) to actual fonts.
Of course, not all 81 combinations of these NFSS weights and widths will map
to existing fonts;
and conversely it may not be possible to assign every existing font
a unique code in a sane way (especially for the weights, some font families
offer more choices or finer granularity than NFSS's codes can handle;
e.g., Fira Sans contains fifteen(!) different weights,
including an additional "Medium" weight between Regular and Semibold).

B<autoinst> tries hard to ensure that the most common NFSS codes
(and high-level commands such as C<\bfseries>,
which are built on top of those codes) will "just work".

To see exactly which NFSS codes map to which fonts, see the log file
(pro tip: run B<autoinst> with the I<-dryrun> option
to check the chosen mapping beforehand).
The I<-nfssweight> and I<-nfsswidth> command-line options can be used
to finetune the mapping between NFSS codes and fonts.

To access specific weights or widths,
one can always use the C<\fontseries> command with the full series name
(i.e., C<\fontseries{demibold}\selectfont>).


=head1 COMMAND-LINE OPTIONS

B<autoinst> tries hard to do The Right Thing (TM) by default,
so you usually won't really need these options;
but most aspects of its operation can be fine-tuned if you want to.

You may use either one or two dashes before options,
and option names may be shortened to a unique prefix
(e.g., B<-encoding> may be abbreviated to B<-enc> or even B<-en>,
but B<-e> is ambiguous (it may mean either B<-encoding> or B<-extra>)).


=over 4

=item B<-version>

Print B<autoinst>'s version number and exit.

=item B<-help>

Print a (relatively) short help text and exit.

=item B<-dryrun>

Don't generate output; just parse input fonts and write
a log file saying what B<autoinst> would have done.

=item B<-logfile>=I<filename>

Write log data to F<filename> instead of the default F<< <fontfamily>.log >>.
If the file already exists, B<autoinst> appends to it;
it doesn't overwrite an existing file.

=item B<-verbose>

Add more details to the log file. Repeat this option for even more info.

=item B<-encoding>=I<encoding[,encoding]>

Generate the specified encoding(s) for the text fonts.
Multiple encodings may be specified as a comma-separated list:
C<-encoding=OT1,LY1,T1> (without spaces!).
The style file passes these to F<otftotfm> in the specified order,
so the I<last> one will become the default text encoding of your document.

The default choice of encodings is "OT1,LY1,T1".
For each encoding, a file F<< <encoding>.enc >> (in all I<lowercase>!)
should be somewhere where F<otftotfm> can find it. Suitable encoding files
for OT1, T1/TS1, LY1, LGR, T2A/B/C and T3/TS3 come with B<autoinst>.
(These files are called F<fontools_ot1.enc> etc. to avoid name clashes
with other packages; the "fontools_" prefix may be omitted.)

=item B<-ts1>/B<-nots1>

Control the creation of TS1-encoded fonts. The default is B<-ts1>
if the text encodings (see I<-encoding> above) include T1,
B<-nots1> otherwise.

=item B<-serif>/B<-sanserif>/B<-typewriter>

Install the font as a serif, sanserif or typewriter font, respectively.
This changes how you access the font in LaTeX:
with C<\rmfamily>/C<\textrm>, C<\sffamily>/C<\textsf>
or C<\ttfamily>/C<\texttt>.

Installing the font as a typewriter font will cause two further changes:
it will - by default - turn off the use of f-ligatures
(though this can be overridden with the I<-ligatures> option),
and it will disable hyphenation for this font.
This latter effect cannot be disabled in B<autoinst>;
if you want typewriter text to be hyphenated, use the F<hyphenat> package.

If none of these options is specified, B<autoinst> tries to guess:
if the font's filename contains the string "mono"
or if the field C<isFixedPitch> in the font's I<post> table is True,
it will select B<-typewriter>;
else if the filename contains "sans" it selects B<-sanserif>;
and otherwise it will opt for B<-serif>.

=item B<-lining>/B<-nolining>

Control the creation of fonts with lining figures. The default is
B<-lining>.

=item B<-oldstyle>/B<-nooldstyle>

Control the creation of fonts with oldstyle figures. The default is
B<-oldstyle>.

=item B<-proportional>/B<-noproportional>

Control the creation of fonts with proportional figures. The default is
B<-proportional>.

=item B<-tabular>/B<-notabular>

Control the creation of fonts with tabular figures. The default is
B<-tabular>.

=item B<-smallcaps>/B<-nosmallcaps>

Control the creation of small caps fonts. The default is
B<-smallcaps>.

=item B<-swash>/B<-noswash>

Control the creation of swash fonts. The default is B<-swash>.

=item B<-titling>/B<-notitling>

Control the creation of titling families. The default is B<-titling>.

=item B<-superiors>/B<-nosuperiors>

Control the creation of fonts with superior characters.
The default is B<-superiors>.

=item B<-noinferiors>

=item B<-inferiors> [= B<none> | B<auto> | B<subs> | B<sinf> | B<dnom> ]

The OpenType standard defines several kinds of digits that might be used
as inferiors or subscripts: "Subscripts" (OpenType feature "subs"),
"Scientific Inferiors" ("sinf"), and "Denominators" ("dnom").
This option allows the user to determine which of these styles B<autoinst>
should use for the inferior characters.
Alternatively, the value "auto" tells B<autoinst> to use the first value
in "subs", "sinf" or "dnom" that is supported by the font.
Saying just B<-inferiors> is equivalent to B<-inferiors=auto>;
otherwise the default is B<-noinferiors>.

I<< If you specify a style of inferiors that isn't present in the font,
B<autoinst> will fall back to its default behaviour of not creating fonts
with inferiors at all; it won't try to substitute one of the other styles. >>

=item B<-fractions>/B<-nofractions>

Control the creation of fonts with numerators and denominators.
The default is B<-nofractions>.

=item B<-ligatures>/B<-noligatures>

Some fonts create glyphs for the standard f-ligatures (ff, fi, fl, ffi, ffl),
but don't provide a "liga" feature to access these.
This option tells B<autoinst> to add extra C<LIGKERN> rules to
the generated fonts to enable the use of these ligatures.
The default is B<-ligatures>,
unless the user specified the I<-typewriter> option.

Specify B<-noligatures> to disable the generation of ligatures even for fonts
that do contain a "liga" feature.

=item B<-defaultlining>/B<-defaultoldstyle>

=item B<-defaulttabular>/B<-defaultproportional>

Tell B<autoinst> which figure style is the current font family's default
(i.e., which figures you get when you don't specify any OpenType features).

I<Don't use these options unless you are certain you need them!>
They are only needed for fonts that don't provide OpenType features
for their default figure style; and even in that case,
B<autoinst>'s default values (B<-defaultlining> and B<-defaulttabular>)
are usually correct.

=item B<-nofigurekern>

Some fonts provide kerning pairs for tabular figures.
This is very probably not what you want
(e.g., numbers in tables won't line up exactly).
This option adds extra I< --ligkern> options
to the commands for I<otftotfm> to suppress such kerns.
Note that this option leads to very long commands (it adds
one hundred I< --ligkern> options), which may cause problems on some systems.

=item B<-mergewidths>/B<-nomergewidths>,
B<-mergeweights>/B<-nomergeweights>,
B<-mergeshapes>/B<-nomergeshapes>

Some font put different widths, weights or shapes (e.g., small caps)
in separate families.
These options tell B<autoinst> to merge those separate families into
the main family.
Since this is usually desirable, they are all enabled by default.

In earlier versions, B<-mergeshapes> was called B<-mergesmallcaps>;
for reasons of backward compatibility, that option is still supported.

=item B<-nfssweight>=I<code>=I<weight>, B<-nfsswidth>=I<code>=I<width>

Map the NFSS code I<code> to the given weight or width,
overriding the built-in tables.
Each of these options may be given multiple times,
to override more than one NFSS code.
Example: to map the "ul" code to the "Thin" weight,
use C<-nfssweight=ul=thin>.
To inhibit the use of the "ul" code completely,
use C<-nfssweight=ul=>.

=item B<-extra>=I<text>

Append I<text> as extra options to the command lines for I<otftotfm>.
To prevent I<text> from accidentily being interpreted as options
to B<autoinst>, it should be properly quoted.

=item B<-manual>

Manual mode; for users who want to post-process the generated files
and commands. By default, B<autoinst> immediately executes all
F<otftotfm> commands it generates;
in manual mode, these are instead written to a file F<autoinst.bat>.
Furthermore it tells F<otftotfm> to generate human readable (and editable)
F<pl/vpl> files instead of the default F<tfm/vf> ones,
and to place all generated files in a subdirectory C<./autoinst_output/>
of the current directory, rather than install them into your TeX installation.

When using this option, you need to execute the following manual steps after
B<autoinst> has finished:

=over 2

=item - run F<pltotf> and F<vptovf> on the generated F<pl> and F<vf> files,
to convert them to F<tfm/vf> format;

=item - move all generated files to a proper TEXMF tree,
and, if necessary, update the filename database;

=item - tell TeX about the new F<map> file
(usually by running C<updmap> or similar).

=back

Note that some options (I<-target>, I<-vendor> and I<-typeface>,
I<-[no]updmap>) are meaningless, and hence ignored, in manual mode.

=item B<-target>=I<DIRECTORY>

Install all generated files into the TEXMF tree at I<DIRECTORY>.

By default, B<autoinst> searches the $TEXMFLOCAL and $TEXMFHOME trees
and installs all files into the first user-writable TEXMF tree it finds.
If B<autoinst> cannot find such a user-writable directory
(which shouldn't happen, since $TEXMFHOME is supposed to be user-writable)
it will print a warning message and put all files into the subdirectory
C<./autoinst_output/> of the current directory.
It's then up to the user to move the generated files to a better location
and update all relevant databases
(usually by calling F<texhash> and F<updmap>).

I<WARNING:> using this option may interfere with F<kpathsea> and F<updmap>
(especially when the chosen directory is outside the standard TEXMF trees),
so using I<-target> will disable the automatic call to F<updmap>
(as if I<-noupdmap> had been given).
It is up to the user to manually update all databases (i.e., by calling
F<texhash> and F<updmap> or similar).

=item B<-vendor>=I<VENDOR>

=item B<-typeface>=I<TYPEFACE>

These options are equivalent to F<otftotfm>'s I< --vendor> and I< --typeface>
options: they change the "vendor" and "typeface" parts of the names of the
subdirectories in the TEXMF tree where generated files will be stored.
The default values are "lcdftools" and the font's FontFamily name.

Note that these options change I<only> directory names,
not the names of any generated files.

=item B<-updmap>/B<-noupdmap>

Control whether or not F<updmap> is called after the last call to F<otftotfm>.
The default is B<-updmap>.

=back


=head1 SEE ALSO

Eddie Kohler's B<TypeTools> (F<http://www.lcdf.org/type>).

B<Perl> can be obtained from F<http://www.perl.org>;
it is included in most Linux distributions.
For Windows, try ActivePerl (F<http://www.activestate.com>)
or Strawberry Perl (F<http://strawberryperl.com>).

B<XeTeX> (F<http://www.tug.org/xetex>) and
B<LuaTeX> (F<http://www.luatex.org>)
are Unicode-aware TeX engines that can use OpenType fonts directly,
without any (La)TeX-specific support files.

The B<FontPro> project (F<https://github.com/sebschub/FontPro>)
offers very complete LaTeX support (even for typesetting maths) for
Adobe's Minion Pro, Myriad Pro and Cronos Pro font families.


=head1 AUTHOR

Marc Penninga (F<marcpenninga@gmail.com>)

When sending a bug report, please give as much relevant information as
possible; this usually includes (but may not be limited to) the log file
(please add the I<-verbose> command-line option, for extra info).
If you see any error messages, please include these I<verbatim>;
don't paraphase.


=head1 COPYRIGHT

Copyright (C) 2005-2020 Marc Penninga.


=head1 LICENSE

This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published
by the Free Software Foundation, either version 2 of the License,
or (at your option) any later version.
A copy of the text of the GNU General Public License is included in
the I<fontools> distribution; see the file F<GPLv2.txt>.


=head1 DISCLAIMER

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.


=head1 VERSION

This document describes B<autoinst> version 20200129.


=head1 RECENT CHANGES

(See the source for the full story, all the way back to 2005.)

=over 12

=item I<2020-01-29>

Don't create empty subdirectories in the target TEXMF tree.

=item I<2019-11-18>

Fine-tuned calling of F<kpsewhich> on Windows (patch by Akira Kakuto).
The font info parsing now also recognises numerical weights, e.g. in Museo.

=item I<2019-10-29>

The generated style files now use T1 as the default text encoding.

=item I<2019-10-27>

The mapping in F<fd> files between font series and standard NFSS attributes
now uses the new I<alias> function instead of I<ssub> (based on code by
Frank Mittelbach).
The way F<otftotfm> is called was changed to work around a Perl/Windows bug;
the old way might cause the process to hang.
Using the I<-target> option now implies I<-noupdmap>, since choosing
a non-standard target directory interferes with kpathsea/texhash and updmap.

=item I<2019-10-01>

Handle I<-target> directories with spaces in their path names.
Tweaked messages and logs to make them more useful to the user.

=item I<2019-07-12>

Replaced single quotes in calls to F<otfinfo> with double quotes,
as they caused problems on Windows 10.

=item I<2019-06-25>

=over 3

=item -

Added the I<-mergeweights> and I<-mergeshapes> options,
and improved I<-mergewidths>.

=item -

Improved the parsing of fonts' widths and weights.

=item -

Improved the mapping of widths and weights to NFSS codes.

=item -

Changed logging code so that that results of font info parsing
are always logged, even (especially!) when parsing fails.

=item -

Added a warning when installing fonts from multiple families.

=item -

Added simple recognition for sanserif and typewriter fonts.

=item -

Fixed error checking after calls to F<otfinfo>
(B<autoinst> previously only checked whether C<fork()> was successful,
not whether the actual call to F<otfinfo> worked).

=item -

Fixed a bug in the I<-inferiors> option;
when used without a (supposedly optional) value,
it would silently gobble the next option instead.

=back

=item I<2019-05-22>

Added the I<mainfont> option to the generated F<sty> files.
Prevented hyphenation for typewriter fonts
(added C<\hyphenchar\font=-1> to the C<\DeclareFontFamily> declarations).
Added the I<-version> option.

=item I<2019-05-17>

Changed the way the F<-ligatures> option works:
F<-ligatures> enables f-ligatures (even without a "liga" feature),
F<-noligatures> now disables f-ligatures (overriding a "liga" feature).

=item I<2019-05-11>

Separate small caps families are now also recognised when the family name
ends with "SC" (previously B<autoinst> only looked for "SmallCaps").

=item I<2019-04-22>

Fixed a bug in the generation of swash shapes.

=item I<2019-04-19>

Fixed a bug that affected -mergesmallcaps with multiple encodings.

=item I<2019-04-16>

Added the <-mergesmallcaps> option, to handle cases where
the small caps fonts are in separate font families.
Titling shape is now treated as a separate family instead of a distinct shape;
it is generated only for fonts with the 'titl' feature.
Only add f-ligatures to fonts when explicitly asked to (I<-ligatures>).

=item I<2019-04-11>

Tried to make the log file more relevant.
Added the I<-nfssweight> and I<-nfsswidth> options,
and finetuned the automatic mapping between fonts and NFSS codes.
Changed the name of the generated log file to F<< <fontfamily>.log >>,
and revived the I<-logfile> option to allow overriding this choice.
Made I<-mergewidths> the default (instead of I<-nomergewidths>).

=item I<2019-04-01>

Fine-tuned the decision where to put generated files;
in particular, create $TEXMFHOME if it doesn't already exist
and $TEXMFLOCAL isn't user-writable.

In manual mode, or when we can't find a user-writable TEXMF tree,
put all generated files into a subdirectory C<./autoinst_output/>
instead of all over the current working directory.

Added "auto" value to the I<inferiors> option,
to tell B<autoinst> to use whatever inferior characters are available.

=item I<2019-03-14>

Overhauled the mapping of fonts (more specifically of weights and widths;
the mapping of shapes didn't change) to NFSS codes.
Instead of inventing our own codes to deal with every possible weight
and width out there, we now create "long" codes based on the names
in the font metadata.
Then we add "ssub" rules to the F<fd> files to map the standard NFSS codes
to our fancy names (see the section B<NFSS codes>;
based on discussions with Frank Mittelbach and Bob Tennent).

=back


=begin Really_old_history

=over 12

=item I<2018-08-10>

Added encoding files for LGR and T2A/B/C to I<fontools>.

=item I<2018-03-26>

Added the I<-(no)mergewidths> option; tried to improve the documentation.

=item I<2018-03-26>

Added the "Text" weight and the I<-(no)mergewidths> option.
Changed the NFSS codes for "Thin" and "Book" to "i" and "o", respectively.
Tried to improve the documentation.

=item I<2018-01-09>

Added the "sl" weight for font families (such as Fira Sans) that contain both
"Book" and "Regular" weights (reported by Bob Tennent).
Added the "Two", "Four", "Eight" and "Hair" weights (for Fira Sans).

=item I<2017-06-16>

Changed the I<-inferiors> option from a binary yes-or-no choice to allow
the user to choose one of the "subs", "sinf" and "dnom" features.
B<autoinst> now always creates a log file.

=item I<2017-03-21>

Updated the F<fontools_ot1.enc> encoding file to include the "Lslash"
and "lslash" glyphs (thanks to Bob Tennent).

=item I<2015-11-22>

Bugfix: LaTeX doesn't like command names with dashes in it.

=item I<2015-05-13>

Fixed an error message that mixed up width and weight.

=item I<2014-04-04>

Fixed a bug in the font info parsing code.

=item I<2014-01-21>

"Oblique" or "slanted" fonts are now mapped to NFSS code "sl" instead
of "it"; added "ssub" rules to the F<fd> files to substitute slanted fonts
for italic ones if the latter are missing. Fixed a few bugs.

=item I<2014-01-03>

Added the I<-dryrun> and I<-logfile> options; changed which info is logged.
Added the I<-lining>, I<-oldstyle>, I<-tabular> and I<-proportional>
options; the old options with those names have been renamed to
I<-defaultlining>, I<-defaultoldstyle> etc.

=item I<2013-10-31>

The previous change required Perl v5.14 or newer;
now it also works with older versions.

=item I<2013-10-01>

Added the I<-lining>, I<-oldstyle>, I<-tabular> and I<-proportional>
command line options.

=item I<2013-07-25>

The generated F<sty> files now use the I<mweights> package instead of
redefining C<\mddefault> and C<\bfdefault>.
Added C<scale> as an alias for the package option C<scaled>.

=item I<2013-02-06>

Bugfix: the directory names for map and encoding files contained
the "vendor" instead of the "typeface".

=item I<2013-01-03>

Added extra "ssub" rules to the F<fd> files that substitute "b" for "bx".
Verbose mode now also prints all generated commands before they're executed.

=item I<2012-10-25>

Added extra "ssub" rules to the F<fd> files that substitute italic
shapes for slanted ones.

=item I<2012-09-25>

Added the I<-vendor>, I<-typeface> and I<-(no)updmap> command line options.

=item I<2012-07-06>

Documentation update.

=item I<2012-03-06>

Implemented the "splitting the font family into multiple subfamilies"
emergency strategy when font info parsing fails.
Added recognition for a number of unusual widths, weights and shapes.

=item I<2012-02-29>

Fixed a bug in the font parsing code,
where possible widths, weights and shapes where tested in the wrong order;
this led to "ExtraLight" fonts being recognised as "Light".
Added recognition for "Narrow" and "Wide" widths.
Also added the I<-(no)figurekern> command-line option.

=item I<2012-02-01>

Reorganised the code, and fixed some bugs in the process.
Added the I<-target> command-line option.
Made B<autoinst> install the F<fd> and F<sty> files in
the same TEXMF tree as the other generated files.
Generate OT1, T1 and LY1 encoded text fonts by default.
Made I<-titling> a default option (instead of I<-notitling>).
Updated the documentation.

=item I<2011-06-15>

Fixed the font info parsing code for some fonts that are
too lazy to spell out "Italic" in full.

=item I<2010-04-29>

The I< --feature=kern> option is only used for fonts that
contain either a I<kern> feature or a I<kern> table.
Font feature selection commands in the F<sty> file are only
generated for shapes and figure styles that are supported
by the current font, and whose generation has not been
turned off using the command-line options.
Fixed the font info parsing to work with the Condensed fonts
in the Minion Pro family.

=item I<2010-04-23>

Always provide the I< --feature=kern> option to F<otftotfm>,
even if the font doesn't have a I<kern> feature;
this will make F<otftotfm> use the I<kern> table if present.
For fonts without a I<liga> feature, add I< --ligkern>
options for the common f-ligatures to the F<otftotfm> command line,
so that any ligatures present in the font will still be used.
Bug fix: the generated F<sty> files now work for font families
with names containing digits.

=item I<2009-04-09>

Prefixed the filenames of  the included encoding files with
I<fontools_>, to prevent name clashes with other packages.

=item I<2009-04-06>

A small patch to the C<get_orn> subroutine:
it now also recognises the I<bullet.xxx> ornament glyphs
in Adobe Kepler Pro.

=item I<2007-08-07>

Fixed a small bug with filename quoting on Windows.

=item I<2007-07-31>

Changed the tables that map weights and widths to NFSS codes:
in some extended families, different combinations of
weight and width were mapped to the same series.
Added a work-around for incorrect size info in some Kepler fonts.
Fixed a small bug in the generated commands for otftotfm
(sometimes, the "onum" feature was included twice).
Added encoding file for OT1 to the I<fontools> collection.

=item I<2007-07-27>

Two bugfixes: a closing brace was missing in the generated style file,
and the NFSS series was formed as "width plus weight" instead of the reverse.

=item I<2007-06-10>

Bugfix: silently replacing \DeclareOption, \ProcessOptions and
\ExecuteOptions with their counterparts from the xkeyval package
caused problems for some other packages.

=item I<2007-06-04>

Added the I< --no-updmap> option to all generated commands for F<otftotfm>
(except the last); this should yield a significant speed-up for large
families (suggested by Steven E. Harris).
Tweaked the font info parsing to work around a bug in the naming of
some FontFont fonts, where every font is in a family of its own.
Added the "scaled" option (including the loading of F<xkeyval>)
to the generated style file.
Extended the output of the I<-verbose> option.

=item I<2007-02-08>

Yet Another Complete Rewrite.
The code is now much more readable and more flexible;
the program can now handle fonts from different families,
as well as multiple text encodings.
Rewrote the font info parsing code to work for Arno Pro.

=item I<2006-10-11>

The program determines the fonts' weights, widths and shapes by parsing
the output from C<otfinfo --info> instead of the font filename.
This should make B<autoinst> work for non-Adobe fonts.
Filenames with spaces now work as well.

=item I<2006-08-31>

Made the generated style files try to include "fontaxes.sty";
changed the names of the generated fonts and families
(to make the previous change possible);
added command-line options for most font styles and shapes;
tweaked the filename parsing code for Cronos Pro and Gill Sans Pro;
added runtime generation of encoding vectors for ornament fonts
(because GaramondPremier's ornament names differ from other fonts);
changed the NFSS-code for italic small caps and titling to "scit" and "tlit"
(to work with F<fontaxes>);
and edited (and hopefully improved) the documentation.

=item I<2005-10-03>

When creating LY1, T1, OT1 or TS1 encoded fonts, the I<-coding-scheme>
option is added to the commands for F<otftotfm>; this should make the
generated F<pl> and F<vpl> files acceptable to I<fontinst>.
Also elaborated the documentation somewhat and fixed a small bug.

=item I<2005-09-22>

Added check to see if filename parsing succeeded;
updated the filename parsing code to cater for GaramondPremier Pro,
Silentium Pro and some non-Adobe fonts;
added the I<-sanserif> and I<-typewriter> options and hacked the
style files to support using several different font families in one document.

=item I<2005-09-12>

Cleaned up the code (it now runs under the F<strict> and F<warnings> pragmas);
fixed a (rather obscure) bug that occurred when creating TS1-encoded
fonts for families with multiple optical masters and oldstyle digits;
added the I<medium, semibold> etc. options to the style file;
and improved the layout of the generated files.

=item I<2005-08-11>

The generated commands weren't actually executed, only printed...
Also added a small hack to cater for fonts
(such as some recent versions of MinionPro)
that contain swash characters but don't provide a "swsh" feature.

=item I<2005-08-10>

Dropped the "fontname" scheme in favor of a more verbose naming scheme,
since many filenames were still more than eight characters long anyway.
Added F<nfssext.sty>-like commands to the generated style file.
Changed the default encoding to LY1 and added the "inferior" shape.

=item I<2005-08-01>

Rewrote (and hopefully improved) the user interface;
changed the program to by default execute the generated F<otftotfm> command
lines rather than writing them to a file;
added automatic determination of the "fontname" code for the font family;
changed the NFSS code for italic small caps to "si"; added titling shapes;
changed the generated style
file to include an interface for the ornaments and to load Lehman's NFSS
extensions F<nfssext.sty> if this is installed; corrected the "fontname" codes
for OT1, T1, LY1 and user-specific encodings; extended the output generated by
the I<-verbose> option; and rewrote and extended the documentation.

=item I<2005-06-16>

Did some more finetuning to the filename-parsing code.

=item I<2005-05-31>

Generate correct fontname for OT1-encoded fonts.

=item I<2005-05-18>

Tried to make the filename-parsing code a bit more robust by adding several
weights and widths; changed the error that's displayed when filename parsing
fails; commented the code.

=item I<2005-04-29>

Rewrote large parts of the code (yes it I<was> even worse).

=item I<2005-04-18>

Changed default text-encoding to T1, added TS1.

=item I<2005-03-29>

Added support for font families with multiple widths.

=item I<2005-03-15>

First version.

=back

=end Really_old_history