path: root/macros/optex/base/table.opm

%% This is part of the OpTeX project, see http://petr.olsak.net/optex

\_codedecl \table {Basic macros for OpTeX <2023-05-19>} % preloaded in format

   \_doc -----------------------------
   The result of the \`\table``{<declaration>}{<data>}` macro is inserted into
   \`\_tablebox`. You can change default value if you want by
   `\let\_tablebox=\vtop` or `\let\_tablebox=\relax`.
   \_cod -----------------------------

\_let\_tablebox=\_vbox

   \_doc -----------------------------
   We save the `to<size>` or `pxto<size>` to `#1` and \`\_tableW` sets the
   `to<size>` to the \`\_tablew` macro.
   If `pxto<size>` is used then `\_tablew` is empty and `\_tmpdim` includes
   given <size>. The `\_ifpxto` returns true in this case.

   The `\table` continues by reading `{<declaration>}` in the \^`\_tableA` macro.
   Catcodes (for example the `|` character) have to be normal when reading
   `\table` parameters. This is the reason why we use `\catcodetable` here.
   \_cod -----------------------------

\_newifi \_ifpxto
\_def\_table#1#{\_tablebox\_bgroup \_tableW#1\_empty\_fin
   \_bgroup \_catcodetable\_optexcatcodes \_tableA}
\_def\_tableW#1#2\_fin{\_pxtofalse
   \_ifx#1\_empty \_def\_tablew{}\_else
   \_ifx#1p \_def\_tablew{}\_tableWx#2\_fin \_else \_def\_tablew{#1#2}\_fi\_fi}
\_def\_tableWx xto#1\_fin{\_tmpdim=#1\_relax \_pxtotrue}
\_public \table ;

   \_doc -----------------------------
   The \^`\tablinespace` is implemented by enlarging given \^`\tabstrut`
   by desired dimension (height and depth too) and by setting
   `\_lineskip=-2\_tablinespace`. Normal table rows (where no `\hrule` is
   between them) have normal baseline distance.\nl
   The \`\_tableA``{<declaration>}` macro scans the `<declaration>` by
   `\_scantabdata#1\_relax` and continues by processing `{<data>}` by \^`\_tableB`.
   The trick `\_tmptoks={<data>}\_edef\_tmpb{\_the\_tmptoks}`
   is used here in order to keep the hash marks in the <data> unchanged.
   \_cod -----------------------------

\_def\_tableA#1{\_egroup
   \_the\_thistable \_global\_thistable={}%
   \_ea\_ifx\_ea^\_the\_tabstrut^\_setbox\_tstrutbox=\_null
   \_else \_setbox\_tstrutbox=\_hbox{\_the\_tabstrut}%
          \_setbox\_tstrutbox=\_hbox{\_vrule width\_zo
              height\_dimexpr\_ht\_tstrutbox+\_tablinespace
              depth\_dimexpr\_dp\_tstrutbox+\_tablinespace}%
          \_offinterlineskip
          \_lineskip=-2\_tablinespace
   \_fi
   \_colnum=0 \_let\_addtabitem=\_addtabitemx
   \_def\_tmpa{}\_tabdata={\_colnum1\_relax}\_scantabdata#1\_relax
   \_the\_everytable \_bgroup \_catcode`\#=12 \_tableB
}

   \_doc -----------------------------
   The \`\_tableB` saves `<data>` to `\_tmpb`
   and does \^`\replstring`s to prefix each macro \^`\crl` (etc.)
   by `\_crcr`. See \^`\_tabreplstrings`.
   The whole `\_tableB` macro is hidden in `{...}` in order to
   there may be `\table` in `\table` and we want to manipulate with
   `&` and `\cr` as with normal tokens in the `\_tabreplstrings`, not as
   the item delimiters of an outer `\table`.

   The `\tabskip` value is saved for places between columns
   into the \`\_tabskipmid` macro. Then it runs
   \begtt \catcode`\<=13
   \tabskip=\tabskipl \halign{<converted declaration>\tabskip=\tabskipr \cr <data>\crcr}
   \endtt
   This sets the desired boundary values of `\tabskip`. The
   \"between-columns" values are set as `\tabskip=`\^`\_tabskipmid` in the
   `<converted declaration>` immediately after each column declarator.

   If `pxto` keyword was used, then we set the virtual unit \^`\tsize` to
   `-\hsize` first. Then the first attempt of the table is created in box 0.
   All collums where `p{..\tsize}` is used, are created as empty in this first
   pass. So, the `\wd0` is the width of all other columns.
   The \^`\_tsizesum` includes the sum of \^`\tsize`'s in `\hsize` units
   after firts pass. The desired table width is stored in the `\_tmpdim`, so
   `\_tmpdim-\_wd0` is the rest which have to be filled by \^`\tsize`s.
   Then the \^`\tsize` is re-calculated and the real table is
   printed by `\halign` in the second pass.

   If no `pxto` keyword was used, then we print the table using `\halign`
   directly. The \^`\_tablew` macro is nonempty if the `to` keyword was used.

   The <data> are re-tokenized by `\_scantextokens` in order to be more
   robust to catcode changing inside the <data>. But inline verbatim cannot
   work in special cases here like \code{`\{`} for example.
   \_cod -----------------------------

\_long\_def\_tableB #1{\_egroup
   {\_def\_tmpb{#1}\_tablereplstrings
    \_edef\_tabskipmid{\_the\_tabskip}\_tabskip=\_tabskipl
    \_ifpxto
      \_edef\_tsizes{\_global\_tsizesum=\_the\_tsizesum \_gdef\_noexpand\_tsizelast{\_tsizelast}}%
      \_tsizesum=\_zo \_def\_tsizelast{0}%
      \_tsize=-\_hsize \_setbox0=\_vbox{\_tablepxpreset \_halign \_tableC}%
      \_advance\_tmpdim by-\_wd0
      \_ifdim \_tmpdim >\_zo \_else \_tsizesum=\_zo \_fi
      \_ifdim \_tsizesum >\_zo \_tsize =\_expr{\_number\_hsize/\_number\_tsizesum}\_tmpdim
      \_else \_tsize=\_zo \_fi
      \_tsizes % retoring values if there is a \table pxto inside a \table pxto.
      \_setbox0=\_null \_halign \_tableC
    \_else
      \_halign\_tablew \_tableC
    \_fi
   }\_egroup % \_tablebox\_bgroup is in the \_table macro
}
\_def\_tableC{\_ea{\_the\_tabdata\_tabskip=\_tabskipr\_cr \_scantextokens\_ea{\_tmpb\_crcr}}}

   \_doc -----------------------------
   \`\_tabreplstrings` replaces each `\crl` etc. to `\crcr\crl`.
   The reason is: we want to use macros that scan its parameter
   to a delimiter written in the right part of the table item declaration.
   The `\crcr` cannot be hidden in another macro in this case.
   \_cod -----------------------------

\_def\_tablereplstrings{%
   \_replstring\_tmpb{\crl}{\_crcr\crl}\_replstring\_tmpb{\crll}{\_crcr\crll}%
   \_replstring\_tmpb{\crli}{\_crcr\crli}\_replstring\_tmpb{\crlli}{\_crcr\crlli}%
   \_replstring\_tmpb{\crlp}{\_crcr\crlp}%
}

\_def\_tablepxpreset{} % can be used to de-activate references to .ref file
\_newbox\_tstrutbox    % strut used in table rows
\_newtoks\_tabdata     % the \halign declaration line

   \_doc -----------------------------
   The \`\_scantabdata` macro converts `\table`'s `<declaration>` to
   `\halign` `<converted declaration>`.
   The result is stored into \`\_tabdata` tokens list.
   For example, the
   following result is generated when `<declaration>=|cr||cl|`.
   \begtt
   tabdata: \_vrule\_the\_tabiteml{\_hfil#\_unsskip\_hfil}\_the\_tabitemr\_tabstrutA
      &\_the\_tabiteml{\_hfil#\_unsskip}\_the\_tabitemr
                                             \_vrule\_kern\_vvkern\_vrule\_tabstrutA
      &\_the\_tabiteml{\_hfil#\_unsskip\_hfil}\_the\_tabitemr\_tabstrutA
      &\_the\_tabiteml{\_relax#\_unsskip\_hfil}\_the\_tabitemr\_vrule\_tabstrutA
   ddlinedata: &\_dditem &\_dditem\_vvitem &\_dditem &\_dditem
   \endtt
   The second result in the \`\_ddlinedata` macro is a template of one row of the table
   used by \^`\crli` macro.
   \_cod -----------------------------

\_def\_scantabdata#1{\_let\_next=\_scantabdata
   \_ifx\_relax#1\_let\_next=\_relax
   \_else\_ifx|#1\_addtabvrule
       \_else\_ifx(#1\_def\_next{\_scantabdataE}%
           \_else\_ifx:#1\_def\_next{\_scantabdataF}%
               \_else\_isinlist{123456789}#1\_iftrue \_def\_next{\_scantabdataC#1}%
                   \_else \_ea\_ifx\_csname _tabdeclare#1\_endcsname \_relax
                       \_ea\_ifx\_csname _paramtabdeclare#1\_endcsname \_relax
                          \_opwarning{tab-declarator "#1" unknown, ignored}%
                       \_else
                          \_def\_next{\_ea\_scantabdataB\_csname _paramtabdeclare#1\_endcsname}\_fi
                       \_else \_def\_next{\_ea\_scantabdataA\_csname _tabdeclare#1\_endcsname}%
   \_fi\_fi\_fi\_fi\_fi\_fi \_next
}
\_def\_scantabdataA#1{\_addtabitem
   \_ea\_addtabdata\_ea{#1\_tabstrutA \_tabskip\_tabskipmid\_relax}\_scantabdata}
\_def\_scantabdataB#1#2{\_addtabitem
   \_ea\_addtabdata\_ea{#1{#2}\_tabstrutA \_tabskip\_tabskipmid\_relax}\_scantabdata}
\_def\_scantabdataC {\_def\_tmpb{}\_afterassignment\_scantabdataD \_tmpnum=}
\_def\_scantabdataD#1{\_loop \_ifnum\_tmpnum>0 \_advance\_tmpnum by-1 \_addto\_tmpb{#1}\_repeat
   \_ea\_scantabdata\_tmpb}
\_def\_scantabdataE#1){\_addtabdata{#1}\_scantabdata}
\_def\_scantabdataF {\_addtabitem\_def\_addtabitem{\_let\_addtabitem=\_addtabitemx}\_scantabdata}

   \_doc -----------------------------
   The \`\_addtabitemx` adds the boundary code (used between columns)
   to the <converted declaration>. This code is
   `\egroup &\bgroup \colnum=<value>\relax`. You can get the current number of
   column from the \`\colnum` register, but you cannot write
   `\the\colnum` as the first object in a <data> item because `\halign`
   first expands the front of the item and the left part of the declaration
   is processed after this. Use `\relax\the\colnum` instead. Or you can
   write:
   \begtt
   \def\showcolnum{\ea\def\ea\totcolnum\ea{\the\colnum}\the\colnum/\totcolnum}
   \table{ccc}{\showcolnum & \showcolnum & \showcolnum}
   \endtt
   This example prints 1/3 \ 2/3 \ 3/3, because the value of the `\colnum` is
   equal to the total number of columns before left part of the column declaration
   is processed.
   \_cod -----------------------------

\_newcount\_colnum     % number of current column in the table
\_public \colnum ;

\_def\_addtabitemx{\_ifnum\_colnum>0
   \_addtabdata{&}\_addto\_ddlinedata{&\_dditem}\_fi
   \_advance\_colnum by1 \_let\_tmpa=\_relax
   \_ifnum\_colnum>1 \_ea\_addtabdata\_ea{\_ea\_colnum\_the\_colnum\_relax}\_fi}
\_def\_addtabdata#1{\_tabdata\_ea{\_the\_tabdata#1}}

   \_doc -----------------------------
   This code converts `||` or `|` from `\table` <declaration> to the <converted declaration>.
   \_cod -----------------------------

\_def\_addtabvrule{%
    \_ifx\_tmpa\_vrule \_addtabdata{\_kern\_vvkern}%
       \_ifnum\_colnum=0 \_addto\_vvleft{\_vvitem}\_else\_addto\_ddlinedata{\_vvitem}\_fi
    \_else \_ifnum\_colnum=0 \_addto\_vvleft{\_vvitemA}\_else\_addto\_ddlinedata{\_vvitemA}\_fi\_fi
    \_let\_tmpa=\_vrule \_addtabdata{\_vrule}%
}
\_def\_tabstrutA{\_copy\_tstrutbox}
\_def\_vvleft{}
\_def\_ddlinedata{}

   \_doc -----------------------------
   The default \"declaration letters" `c`, `l`, `r` and `p` are declared
   by setting \`\_tabdeclarec`, \`\_tabdeclarel`, \`\_tabdeclarer` and
   \^`\_paramtabdeclarep` macros. In general, define
   `\def\_tabdeclare<letter>{...}` for a non-parametric
   letter and `\def\_paramtabdeclare<letter>{...}` for a letter with a parameter.
   The double hash `##` must be in the definition, it is replaced by a real table item data.
   You can declare more such \"declaration letters" if you want.

   Note, that the `##` with fills are in group. The reason can be explained
   by following example:
   \begtt
   \table{|c|c|}{\crl \Red A & B \crl}
   \endtt
   We don't want vertical line after red A to be in red.
   \_cod -----------------------------

\_def\_tabdeclarec{\_the\_tabiteml \_hfil{##}\_unsskip \_hfil \_the\_tabitemr}
\_def\_tabdeclarel{\_the\_tabiteml {##}\_unsskip \_hfil\_the\_tabitemr}
\_def\_tabdeclarer{\_the\_tabiteml \_hfil{##}\_unsskip \_the\_tabitemr}

   \_doc -----------------------------
   The \`\_paramtabdeclarep``{<data>}` is invoked when `p{<data>}`
   declarator is used. First, it saves the `\hsize` value
   and then it runs \`\_tablepar`.
   The \^`\_tablepar` macro behaves like \`\_tableparbox` (which is `\vtop`) in normal cases.
   But there is a special case: if the first pass of `pxto` table is
   processed then `\hsize` is negative. We print nothing in this case, i.e.\
   \^`\_tableparbox` is \^`\ignoreit` and we advance the \`\_tsizesum`.
   The auxiliary macro \`\_tsizelast` is used to do advancing
   only in the first row of the table.
   \^`\_tsizesum` and \^`\_tsizelast` are initialized in the \^`\_tableB` macro.
   \_cod -----------------------------

\_def\_paramtabdeclarep#1{\_hsize=#1\_relax
   \_the\_tabiteml \_tablepar{\_tableparB ##\_tableparC}\_the\_tabitemr
}
\_def\_tablepar{%
   \_ifdim\_hsize<0pt
      \_ifnum\_tsizelast<\_colnum \_global\_advance\_tsizesum by-\_hsize
         \_xdef\_tsizelast{\_the\_colnum}\_fi
      \_let\_tableparbox=\_ignoreit
   \_fi
   \_tableparA \_tableparbox
}
\_let \_tableparbox=\_vtop
\_let \_tableparA=\_empty
\_newdimen \_tsizesum
\_def \_tsizelast{0}

   \_doc -----------------------------
   The \`\_tableparB` initializes the paragraphs inside the table item and
   \`\_tableparC` closes them. They are used in the \^`\_paramtabdeclarep` macro.
   The first paragraph is no indented.
   \_cod -----------------------------

\_def\_tableparB{%
   \_baselineskip=\_normalbaselineskip \_lineskiplimit=\_zo \_noindent
   \_unless\_ifx\_tabstrutA\_empty \_raise\_ht\_tstrutbox\_null \_fi
   \_hskip\_zo \_relax
}
\_def\_tableparC{%
   \_unsskip
   \_unless\_ifx\_tabstrutA\_empty
      \_ifvmode\_vskip\_dp\_tstrutbox \_else\_lower\_dp\_tstrutbox\_null\_fi
   \_fi
}


   \_doc -----------------------------
   Users put optional spaces around the table item typically, i.e.\ they write
   `& text &` instead `&text&`. The left space is ignored by the internal \TeX/ algorithm but
   the right space must be removed by macros. This is a reason why we recommend to
   use \`\_unsskip` after each `##` in your definition of \"declaration letters".
   This macro isn't only the primitive `\unskip` because we allow usage of plain \TeX/
   `\hideskip` macro: `&\hideskip text\hideskip&`.
   \_cod -----------------------------

\_def\_unsskip{\_ifmmode\_else\_ifdim\_lastskip>\_zo \_unskip\_fi\_fi}

   \_doc -----------------------------
   The \`\fL`, \`\fR`, \`\fC` and \`\fX` macros only do special parameters settings
   for paragraph building algorithm.
   \_cod

\_let\_fL=\_raggedright
\_def\_fR{\_leftskip=0pt plus 1fill \_relax}
\_def\_fC{\_leftskip=0pt plus1fill \_rightskip=0pt plus 1fill \_relax}
\_def\_fX{\_leftskip=0pt plus1fil  \_rightskip=0pt plus-1fil \_parfillskip=0pt plus2fil \_relax}
\_public \fL \fR \fC \fX ;

   \_doc -----------------------------
   The \`\fS` macro is more tricky. The \^`\_tableparbox` isn't printed
   immediatelly, but `\setbox2=` is prefixed by the macro \`\_tableparA`,
   which is empty by default (used in \^`\_tablepar`). The \`\_tableparD`
   is processed after the box is set: it checks if there is only one line
   and prints `\hbox to\hsize{\hfil<this line>\hfil}` in this case. In other
   cases, the box2 is printed.
   \_cod -----------------------------

\_def\_fS{\_relax
   \_ifdim\_hsize<0pt \_else \_def\_tableparA{\_setbox2=}\_fi
   \_addto\_tableparC{\_aftergroup\_tableparD}%
}
\_def\_tableparD{\_setbox0=\_vbox{\_unvcopy2 \_unskip \_global\_setbox1=\_lastbox}%
   \_ifdim\_ht0>0pt \_box2 \_setbox0=\_box1
   \_else \_hbox to\_hsize{\_hfil \_unhbox1\_unskip\_unskip\_hfil}\_setbox0=\_box2 \_fi
}
\_public \fS ;

   \_doc -----------------------------
   The family of `\_cr*` macros
   \`\crl`, \`\crll`, \`\crli`, \`\crlli`, \`\crlp` and
   \`\tskip` `<dimen>` is implemented here.
   The \`\_zerotabrule` is used to suppress the negative `\lineskip`
   declared by \^`\tablinespace`.
   \_cod -----------------------------

\_def\_crl{\_crcr\_noalign{\_hrule}}
\_def\_crll{\_crcr\_noalign{\_hrule\_kern\_hhkern\_hrule}}
\_def\_zerotabrule {\_noalign{\_hrule height\_zo width\_zo depth\_zo}}

\_def\_crli{\_crcr \_zerotabrule \_omit
   \_gdef\_dditem{\_omit\_tablinefil}\_gdef\_vvitem{\_kern\_vvkern\_vrule}\_gdef\_vvitemA{\_vrule}%
   \_vvleft\_tablinefil\_ddlinedata\_crcr \_zerotabrule}
\_def\_crlli{\_crli\_noalign{\_kern\_hhkern}\_crli}
\_def\_tablinefil{\_leaders\_hrule\_hfil}

\_def\_crlp#1{\_crcr \_zerotabrule \_noalign{\_kern-\_drulewidth}%
   \_omit \_xdef\_crlplist{#1}\_xdef\_crlplist{,\_ea}\_ea\_crlpA\_crlplist,\_fin,%
   \_global\_tmpnum=0 \_gdef\_dditem{\_omit\_crlpD}%
   \_gdef\_vvitem{\_kern\_vvkern\_kern\_drulewidth}\_gdef\_vvitemA{\_kern\_drulewidth}%
   \_vvleft\_crlpD\_ddlinedata \_global\_tmpnum=0 \_crcr \_zerotabrule}
\_def\_crlpA#1,{\_ifx\_fin#1\_else \_crlpB#1-\_fin,\_ea\_crlpA\_fi}
\_def\_crlpB#1#2-#3,{\_ifx\_fin#3\_xdef\_crlplist{\_crlplist#1#2,}\_else\_crlpC#1#2-#3,\_fi}
\_def\_crlpC#1-#2-#3,{\_tmpnum=#1\_relax
   \_loop \_xdef\_crlplist{\_crlplist\_the\_tmpnum,}\_ifnum\_tmpnum<#2\_advance\_tmpnum by1 \_repeat}
\_def\_crlpD{\_incr\_tmpnum \_edef\_tmpa{\_noexpand\_isinlist\_noexpand\_crlplist{,\_the\_tmpnum,}}%
   \_tmpa\_iftrue \_kern-\_drulewidth \_tablinefil \_kern-\_drulewidth\_else\_hfil \_fi}

\_def\_tskip{\_afterassignment\_tskipA \_tmpdim}
\_def\_tskipA{\_gdef\_dditem{}\_gdef\_vvitem{}\_gdef\_vvitemA{}\_gdef\_tabstrutA{}%
    \_vbox to\_tmpdim{}\_ddlinedata \_crcr
    \_zerotabrule \_noalign{\_gdef\_tabstrutA{\_copy\_tstrutbox}}}

\_public \crl \crll \crli \crlli \crlp \tskip ;

   \_doc -----------------------------
   The \`\mspan``{<number>}[<declaration>]{<text>}` macro generates similar `\omit\span\omit\span`
   sequence as plain \TeX/ macro `\multispan`. Moreover, it uses \^`\_scantabdata` to
   convert `<declaration>` from `\table` syntax to `\halign` syntax.
   \_cod -----------------------------

\_def\_mspan{\_omit \_afterassignment\_mspanA \_mscount=}
\_def\_mspanA[#1]#2{\_loop \_ifnum\_mscount>1 \_cs{_span}\_omit \_advance\_mscount-1 \_repeat
   \_count1=\_colnum \_colnum=0 \_def\_tmpa{}\_tabdata={}\_scantabdata#1\_relax
   \_colnum=\_count1 \_setbox0=\_vbox{\_halign\_ea{\_the\_tabdata\_cr#2\_cr}%
   \_global\_setbox8=\_lastbox}%
   \_setbox0=\_hbox{\_unhbox8 \_unskip \_global\_setbox8=\_lastbox}%
   \_unhbox8 \_ignorespaces}
\_public \mspan ;

   \_doc -----------------------------
   The \`\vspan``<number>{<text>}` implementation is here.
   We need to lower the box by
   \begtt \catcode`\<=13
   (<number>-1)*(\ht+\dp of \tabstrut) / 2.
   \endtt
   The `#1` parameter must be a one-digit number. If you want to set more digits
   then use braces.
   \_cod -----------------------------

\_def\_vspan#1#2#{\_vspanA{#1#2}}
\_def\_vspanA#1#2{\_vtop to\_zo{\_hbox{\_lower \_dimexpr
   #1\_dimexpr(\_ht\_tstrutbox+\_dp\_tstrutbox)/2\_relax
    -\_dimexpr(\_ht\_tstrutbox+\_dp\_tstrutbox)/2\_relax \_hbox{#2}}\_vss}}
\_public \vspan ;

   \_doc -----------------------------
   The parameters of primitive `\vrule` and `\hrule` keeps the rule \"last wins".
   If we re-define `\hrule` to `\_orihrule height1pt` then each usage
   of redefined `\hrule` uses `1pt` height if this parameter isn't
   overwritten by another following `height` parameter. This principle is used for
   settings another default rule thickness than 0.4\,pt by the macro
   \`\rulewidth`.
   \_cod -----------------------------

\_newdimen\_drulewidth  \_drulewidth=0.4pt
\_let\_orihrule=\_hrule  \_let\_orivrule=\_vrule
\_def\_rulewidth{\_afterassignment\_rulewidthA \_drulewidth}
\_def\_rulewidthA{\_edef\_hrule{\_orihrule height\_drulewidth}%
                  \_edef\_vrule{\_orivrule width\_drulewidth}%
                  \_let\_rulewidth=\_drulewidth
                  \_public \vrule \hrule \rulewidth;}
\_public \rulewidth ;

   \_doc -----------------------------
   The \`\frame``{<text>}` uses \"\code{\\vbox} in \code{\\vtop}" trick in order to keep the
   baseline of the internal text at the same level as outer baseline.
   User can write `\frame{abcxyz}` in normal paragraph line, for example
   and gets the expected result: \frame{abcxyz}.
   The internal margins are set by `\vvkern` and `\hhkern` parameters.
   \_cod -----------------------------

\_long\_def\_frame#1{%
   \_hbox{\_vrule\_vtop{\_vbox{\_hrule\_kern\_vvkern
      \_hbox{\_kern\_hhkern\_relax#1\_kern\_hhkern}%
   }\_kern\_vvkern\_hrule}\_vrule}}
\_public \frame ;

   \_doc -----------------------------
   \`\eqbox` and \`\eqboxsize` are implemented here. The widths of
   all `\eqbox`es are saved to the `.ref` file in the format
   \`\_Xeqbox``{<label>}{<size>}`. The `.ref` file is read again and maximum
   box width for each <label> is saved to `\_eqb:<label>`.
   \_cod -----------------------------

\_def\_Xeqbox#1#2{%
   \_ifcsname _eqb:#1\_endcsname
      \_ifdim #2>\_cs{_eqb:#1}\_relax \_sdef{_eqb:#1}{#2}\_fi
   \_else \_sdef{_eqb:#1}{#2}\_fi
}
\_def\_eqbox #1[#2]#3{\_setbox0=\_hbox{{#3}}%
   \_openref \_immediate\_wref \_Xeqbox{{#2}{\_the\_wd0}}%
   \_ifcsname _eqb:#2\_endcsname
       \_hbox to\_cs{_eqb:#2}{\_ifx r#1\_hfill\_fi\_hss\_unhbox0\_hss\_ifx l#1\_hfill\_fi}%
   \_else \_box0 \_fi
}
\_def\_eqboxsize [#1]#2{\_trycs{_eqb:#1}{#2}}

\public \eqbox \eqboxsize ;

\_endcode % -------------------------------------

\secc[table.bound] The boundary declarator \code{:}

The `<declaration>` part of `\table{<declaration>}{<data>}` includes
column declarators (letters) and other material: the `|` or `(<cmd>)`.
If the boundary declarator `:` is not used then
the boundaries of columns are just before each column declarator
with exception of the first one.
For example, the declaration  `{|c||c(xx)(yy)c}`
should be written more exactly using the boundary declarator `:`
by `{|c||:c(xx)(yy):c}`. But you can set these boundaries to
other places using the boundary declarator~`:` explicitly,
for example `{|c:||c(xx):(yy)c}`. The boundary declarator `:`
can be used only once between each pair of column declarators.

Each table item has its group. The `(<cmd>)` are parts of the given
table item (depending on the boundary declarator position).
If you want to apply a special setting for a given column, you can do this by
`(<setting>)` followed by column declarator. But if the column is not
first, you must use `:(<setting>)`. Example. We have three centered columns,
the second one have to be in bold font and the third one have to be in red:
`\table{c:(\bf)c:(\Red)c}{<data>}`

\secc Usage of the \code{\\tabskip} primitive

The value of `\tabskip` primitive is used between all columns of the table.
It is glue-type, so it can be stretchable or shrinkable, see next
section~\ref[table.w].

By default, `\tabskip` is 0\,pt. It means that only \^`\tabiteml`,
\^`\tabitemr` and `(<cmds>)` can generate visual spaces between columns. But they
are not real spaces between columns because they are in fact the part of
the total column width.

The `\tabskip` value declared before the \^`\table` macro (or in
\^`\everytable` or in \^`\thistable`) is used between all columns in the table.
This value is equal to all spaces between columns. But you can set each such space
individually if you use `(\tabskip=<value>)` in the <declaration> immediately
before boundary character. The boundary character represents the column pair for
which the `\tabskip` has individual value. For example
`c(\tabskip=5pt):r` gives `\tabskip` value between `c` and `r` columns.
You need not use boundary character explicitly, so `c(\tabskip=5pt)r` gives
the same result.

Space before the first column is given by the \^`\tabskipl`
and space after the last column is equal to \^`\tabskipr`.
Default values are 0\,pt.

Use nonzero `\tabskip` only in special applications. If `\tabskip` is
nonzero then horizontal lines generated by \^`\crli`, \^`\crlli` and \^`\crlp`
have another behavior than you probably expected: they are interrupted in
each `\tabskip` space.

\secc[table.w] Tables to given width

There are two possibilities how to create tables to given width:

\begitems
* `\table to<size>{<declaration>}{<data>}` uses stretchability or
  shrinkability of all spaces between columns generated by `\tabskip` value and
  eventually by \^`\tabskipl`, \^`\tabskipr` values. See example below.
* `\table pxto<size>{<declaration>}{<data>}` expands the columns declared
  by `p{<size>}`, if the `<size>` is given by a virtual \^`\tsize` unit.
  See the example below.
\enditems

Example of `\table to<size>`:

\begtt \catcode`\<=13
\thistable{\tabskip=0pt plus1fil minus1fil}
\table to\hsize {lr}{<data>}
\endtt
%
This table has its width `\hsize`. The first column starts at the left boundary of
this table and it is justified left (to the boundary). The second column ends at
the right boundary of the table and it is justified right (to the boundary).
The space between them is stretchable and shrinkable to reach
the given width `\hsize`.

\medskip
Example of `\table pxto<size>` (means \"{\bf p}aragraphs e{\bf x}panded {\bf to}"):

\begtt
\table pxto\hsize {|c|p{\tsize}|}{\crl
 aaa         & Ddkas jd dsjds ds cgha sfgs dd fddzf dfhz xxz
               dras ffg hksd kds d sdjds h sd jd dsjds ds cgha
               sfgs dd fddzf dfhz xxz. \crl
 bb ddd ggg  & Dsjds ds cgha sfgs dd fddzf dfhz xxz
               ddkas jd dsjds ds cgha sfgs dd fddzf. \crl }
\endtt

\noindent\table pxto\hsize {|c|p{\tsize}|}{\crl
 aaa         & Ddkas jd dsjds ds cgha sfgs dd fddzf dfhz xxz
               dras ffg hksd kds d sdjds h sd jd dsjds ds cgha
               sfgs dd fddzf dfhz xxz. \crl
 bb ddd ggg  & Dsjds ds cgha sfgs dd fddzf dfhz xxz
               ddkas jd dsjds ds cgha sfgs dd fddzf. \crl }
\medskip

The first `c` column is variable width (it gets the width of the most wide item)
and the resting space to given `\hsize` is filled by the `p` column.

You can declare more than one `p{<coefficient>\tsize}` columns in the table
when `pxto` keyword is used.

\begtt \catcode`\<=13
\table pxto13cm {r p{3.5\tsize} p{2\tsize} p{\tsize} l}{<data>}
\endtt
%
This gives the ratio of widths of individual paragraphs in the table
3.5\,:\,2\,:\,1.

\secc `\eqbox`: boxes with equal width across the whole document

The \^`\eqbox` `[<label>]{<text>}` behaves like `\hbox{<text>}` in the first
run of \TeX. But the widths of all boxes with the same label are saved to
`.ref` file and the maximum box width for each label is calculated at the
beginning of the next \TeX/ run. Then \^`\eqbox`~`[<label>]{<text>}` behaves
like `\hbox to <dim:label> {\hss <text>\hss}`, where <dim:label> is
the maximum width of all boxes
labeled by the same `[<label>]`. The documentation of the \LaTeX/ package
`eqparbox` includes more information and tips.

The \^`\eqboxsize` `[<label>]{<dimen>}` expands to <dim:label> if this value is known,
else it expands to the given <dimen>.

The optional parameter `r` or `l` can be written before `[<label>]` (for example
`\eqbox r[label]{text}`) if you want to put the text to the right or to the
left side of the box width.

Try the following example and watch what happens after first \TeX/ run and
after the second one.

\begtt
\def\leftitem#1{\par
   \noindent \hangindent=\eqboxsize[items]{2em}\hangafter=1
   \eqbox r[items]{#1 }\ignorespaces}

\leftitem {\bf first}      \lorem[1]
\leftitem {\bf second one} \lorem[2]
\leftitem {\bf final}      \lorem[3]
\endtt


\_endinput

2023-05-19 \_tableparB,C: no strut when \tskip is processed (bug fixed).
2022-12-26 \tabdeclare: use {##} in order to allow a\over b in table items
2022-05-03 \_let\_tablereplstring\_relax removed and {...} encloses whole \_tableB body
2022-05-02 \_tablereplstrings put in group: bug fix \table in \halign
2022-02-25 {}->\(begin|end)group in \_tabdeclare(l|r|c), bug when \tabitem(l|r)={$} fixed.
2021-08-04 \_patamtabdeclarep reimplemented.
2021-07-16 grouping based on colors via attributes.
2021-06-03 \scantextokens added to \_partabitem, bug fixed.
2021-06-03 \fS corrected, re-implemented.
2021-06-02 \table in \table allowed, bug fixed.
2021-06-01 pxto: sum of \tsize's allowed different than 1
2021-04-12 p declarator: \hskip0pt added to allow first word hyphenated
2021-03-09 \_tablepxpreset introduced
2021-02-15 \_expandafter -> \_ea
2021-01-27 \_tableB: \long\def added, \vspan syntax improved
2020-05-26 \eqbox implemented
2020-05-20 \colnum implemented.
           \vspan implemented.
2020-05-19 \fL, \fR ets defined.
           \table to, \table pxto implemented.
2020-05-10 \bgroup &\egroup added to tabdata (in order to \aftergroup from colors)
           boundary declarator : introduced
           \_unsskip: \ifmmode added.
2020-04-10 second group for items in order to \localcolors