[latex3-commits] [git/LaTeX3-latex3-latex3] main: Implement regex_case_match (see #433) (5f6e07a24)
Joseph Wright
joseph.wright at morningstar2.co.uk
Mon Jan 10 19:08:09 CET 2022
Repository : https://github.com/latex3/latex3
On branch : main
Link : https://github.com/latex3/latex3/commit/5f6e07a245d47bfa6b28e2904d850ad9638bec6c
>---------------------------------------------------------------
commit 5f6e07a245d47bfa6b28e2904d850ad9638bec6c
Author: Bruno Le Floch <blflatex at gmail.com>
Date: Thu May 6 13:20:41 2021 +0200
Implement regex_case_match (see #433)
>---------------------------------------------------------------
5f6e07a245d47bfa6b28e2904d850ad9638bec6c
l3kernel/l3regex.dtx | 236 ++++++++++++++++++++-
l3kernel/testfiles/m3regex012.lvt | 50 +++++
.../testfiles/{m3clist006.tlg => m3regex012.tlg} | 39 ++--
3 files changed, 299 insertions(+), 26 deletions(-)
diff --git a/l3kernel/l3regex.dtx b/l3kernel/l3regex.dtx
index 12e78e9d9..5f3fe80c2 100644
--- a/l3kernel/l3regex.dtx
+++ b/l3kernel/l3regex.dtx
@@ -99,7 +99,7 @@
%
% \section{Syntax of regular expressions}
%
-% \subsection{Regex examples}
+% \subsection{Regular expression examples}
%
% We start with a few examples, and encourage the reader to apply
% \cs{regex_show:n} to these regular expressions.
@@ -620,6 +620,43 @@
% results in \cs[no-index]{l_foo_int} taking the value $5$.
% \end{function}
%
+% \begin{function}[noTF, added = 2021-05-15]{\regex_case_match:nn}
+% \begin{syntax}
+% \cs{regex_case_match:nnTF}
+% ~~|{| \\
+% ~~~~\Arg{regex_1} \Arg{code case_1} \\
+% ~~~~\Arg{regex_2} \Arg{code case_2} \\
+% ~~~~\ldots \\
+% ~~~~\Arg{regex_n} \Arg{code case_n} \\
+% ~~|}| \Arg{token list}
+% ~~\Arg{true code} \Arg{false code}
+% \end{syntax}
+% Determines which of the \meta{regular expressions} matches at the
+% earliest point in the \meta{token list}, and leaves the
+% corresponding \meta{code_i} followed by the \meta{true code} in the
+% input stream. If several \meta{regex} match starting at the same
+% point, then the first one in the list is selected and the others are
+% discarded. If none of the \meta{regex} match, the \meta{false code}
+% is left in the input stream.
+%
+% In detail, for each starting position in the \meta{token list}, each
+% of the \meta{regex} is searched in turn. If one of them matches
+% then the corresponding \meta{code} is used and everything else is
+% discarded, while if none of the \meta{regex} match at a given
+% position then the next starting position is attempted. If none of
+% the \meta{regex} match anywhere in the \meta{token list} then
+% nothing is left in the input stream. Note that this differs from
+% nested \cs{regex_match:nnTF} statements since all \meta{regex} are
+% attempted at each position rather than attempting to match
+% \meta{regex_1} at every position before moving on to \meta{regex_2}.
+%
+% Each \meta{regex} can either be given as a regex variable or as an
+% explicit regular expression. It may be useful to include as the
+% last case \meta{regex_n} the regex |{\Z}|, which always matches at
+% the very end of the \meta{token list}, so as to include fall-back
+% code \meta{code case_n} used when no other case matches.
+% \end{function}
+%
% \section{Submatch extraction}
%
% \begin{function}[noTF, added = 2017-05-26]
@@ -1100,6 +1137,27 @@
% \end{macrocode}
% \end{macro}
%
+% \begin{macro}{\@@_tl_pairs_map:nNF, \@@_tl_pairs_map:Nnn}
+% Map through a token list one pair at a time, with a two-argument
+% function.
+% \begin{macrocode}
+\cs_new:Npn \@@_tl_pairs_map:nNF #1#2
+ {
+ \@@_tl_pairs_map:Nnn #2 #1 \q_@@_nil \q_@@_nil
+ \prg_break_point:
+ }
+\cs_new:Npn \@@_tl_pairs_map:Nnn #1#2#3
+ {
+ \@@_use_none_delimit_by_q_nil:w
+ #2 \prg_break:n \use_none:n
+ #3 \prg_break:n \use:n
+ \q_@@_nil
+ #1 {#2} {#3}
+ \@@_tl_pairs_map:Nnn #1
+ }
+% \end{macrocode}
+% \end{macro}
+%
% \subsubsection{Constants and variables}
%
% \begin{macro}{\@@_tmp:w}
@@ -1186,9 +1244,17 @@
% \end{macrocode}
% \end{variable}
%
+% \begin{variable}{\q_@@_nil}
+% Internal quarks.
+% \begin{macrocode}
+\quark_new:N \q_@@_nil
+% \end{macrocode}
+% \end{variable}
+%
% \begin{macro}[EXP]{
% \@@_use_none_delimit_by_q_recursion_stop:w,
-% \@@_use_i_delimit_by_q_recursion_stop:nw
+% \@@_use_i_delimit_by_q_recursion_stop:nw,
+% \@@_use_none_delimit_by_q_nil:w,
% }
% Functions to gobble up to a quark.
% \begin{macrocode}
@@ -1196,16 +1262,10 @@
#1 \q_@@_recursion_stop { }
\cs_new:Npn \@@_use_i_delimit_by_q_recursion_stop:nw
#1 #2 \q_@@_recursion_stop {#1}
+\cs_new:Npn \@@_use_none_delimit_by_q_nil:w #1 \q_@@_nil { }
% \end{macrocode}
% \end{macro}
%
-% \begin{variable}{\q_@@_nil}
-% Internal quarks.
-% \begin{macrocode}
-\quark_new:N \q_@@_nil
-% \end{macrocode}
-% \end{variable}
-%
% \begin{macro}[pTF]{\@@_quark_if_nil:n}
% Branching quark conditional.
% \begin{macrocode}
@@ -2343,6 +2403,24 @@
% \end{macrocode}
% \end{macro}
%
+% \begin{macro}{\@@_compile_use:n}
+% Use a regex, regardless of whether it is given as a string (in which
+% case we need to compile) or as a regex variable. This is used for
+% \cs{regex_case_match:nn} and related functions to allow a mixture of
+% explicit regex and regex variables.
+% \begin{macrocode}
+\cs_new_protected:Npn \@@_compile_use:n #1
+ {
+ \tl_if_single_token:nT {#1}
+ {
+ \exp_args:No \tl_if_head_eq_meaning:nNT #1 \@@_branch:n
+ { \use_ii:nnn }
+ }
+ \@@_compile:n {#1} \l_@@_internal_regex
+ }
+% \end{macrocode}
+% \end{macro}
+%
% \begin{macro}{\@@_compile_escaped:N, \@@_compile_special:N}
% If the special character or escaped alphanumeric has a particular
% meaning in regexes, the corresponding function is used. Otherwise,
@@ -4154,6 +4232,76 @@
% \end{macrocode}
% \end{macro}
%
+% \begin{variable}{\g_@@_case_int}
+% Case number that was successfully matched in
+% \cs{regex_case_match:nn} and related functions.
+% \begin{macrocode}
+\int_new:N \g_@@_case_int
+% \end{macrocode}
+% \end{variable}
+%
+% \begin{variable}{\l_@@_case_max_group_int}
+% The largest group number appearing in any of the \meta{regex} in the
+% argument of \cs{regex_case_match:nn} and related functions.
+% \begin{macrocode}
+\int_new:N \l_@@_case_max_group_int
+% \end{macrocode}
+% \end{variable}
+%
+% \begin{macro}{\@@_case_build:n, \@@_case_build_aux:Nn, \@@_case_build_loop:n}
+% See \cs{@@_build:n}, but with a loop.
+% \begin{macrocode}
+\cs_new_protected:Npn \@@_case_build:n
+ { \@@_case_build_aux:Nn \c_true_bool }
+\cs_new_protected:Npn \@@_case_build_aux:Nn #1#2
+ {
+ \@@_standard_escapechar:
+ \int_set_eq:NN \l_@@_max_state_int \l_@@_min_state_int
+ \@@_build_new_state:
+ \@@_build_new_state:
+ \@@_toks_put_right:Nn \l_@@_left_state_int
+ { \@@_action_start_wildcard:N #1 }
+ %
+ \@@_build_new_state:
+ \@@_toks_put_left:Nx \l_@@_left_state_int
+ { \@@_action_submatch:nN { 0 } < }
+ \@@_push_lr_states:
+ \int_zero:N \l_@@_case_max_group_int
+ \int_gzero:N \g_@@_case_int
+ \tl_map_inline:nn {#2}
+ {
+ \int_gincr:N \g_@@_case_int
+ \@@_case_build_loop:n {##1}
+ }
+ \int_set_eq:NN \l_@@_capturing_group_int \l_@@_case_max_group_int
+ \@@_pop_lr_states:
+ }
+\cs_new_protected:Npn \@@_case_build_loop:n #1
+ {
+ \int_set:Nn \l_@@_capturing_group_int { 1 }
+ \@@_compile_use:n {#1}
+ \int_set:Nn \l_@@_case_max_group_int
+ {
+ \int_max:nn { \l_@@_case_max_group_int }
+ { \l_@@_capturing_group_int }
+ }
+ \seq_pop:NN \l_@@_right_state_seq \l_@@_internal_a_tl
+ \int_set:Nn \l_@@_right_state_int \l_@@_internal_a_tl
+ \@@_toks_put_left:Nx \l_@@_right_state_int
+ {
+ \@@_action_submatch:nN { 0 } >
+ \int_gset:Nn \g_@@_case_int
+ { \int_use:N \g_@@_case_int }
+ \@@_action_success:
+ }
+ \@@_toks_clear:N \l_@@_max_state_int
+ \seq_push:No \l_@@_right_state_seq
+ { \int_use:N \l_@@_max_state_int }
+ \int_incr:N \l_@@_max_state_int
+ }
+% \end{macrocode}
+% \end{macro}
+%
% \begin{macro}{\@@_build_for_cs:n}
% The matching code relies on some global intarray variables, but only
% uses a range of their entries. Specifically,
@@ -4495,7 +4643,7 @@
% the group, and leaves \texttt{internal_a} pointing to the left end
% of the last repetition. We only record the submatch information at
% the last repetition. Finally, add a state at the end (the transition
-% to it has been taken care of by the replicating auxiliary.
+% to it has been taken care of by the replicating auxiliary).
% \begin{macrocode}
\cs_new_protected:Npn \@@_group_repeat:nn #1#2
{
@@ -6269,6 +6417,34 @@
% \end{macrocode}
% \end{macro}
%
+% \begin{macro}[noTF]{\regex_case_match:nn}
+% The auxiliary errors if |#1| has an odd number of items, and
+% otherwise it sets \cs{g_@@_case_int} according to which case was
+% found (zero if not found). Overall, the \texttt{false} branch is
+% taken either if the cases |#1| were nonsense (odd number of items)
+% or if there was no match. The \texttt{true} branch is taken
+% otherwise, and we leave the corresponding code in the input stream.
+% \begin{macrocode}
+\cs_new_protected:Npn \regex_case_match:nnTF #1#2#3#4
+ {
+ \int_gzero:N \g_@@_case_int
+ \@@_case_match:nn {#1} {#2}
+ \int_compare:nNnTF \g_@@_case_int = 0
+ {#4}
+ {
+ \tl_item:nn {#1} { 2 * \g_@@_case_int }
+ #3
+ }
+ }
+\cs_new_protected:Npn \regex_case_match:nn #1#2
+ { \regex_case_match:nnTF {#1} {#2} { } { } }
+\cs_new_protected:Npn \regex_case_match:nnT #1#2#3
+ { \regex_case_match:nnTF {#1} {#2} {#3} { } }
+\cs_new_protected:Npn \regex_case_match:nnF #1#2#3
+ { \regex_case_match:nnTF {#1} {#2} { } {#3} }
+% \end{macrocode}
+% \end{macro}
+%
% \begin{macro}[noTF]
% {
% \regex_extract_once:nnN, \regex_extract_once:NnN,
@@ -6433,6 +6609,37 @@
% \end{macrocode}
% \end{macro}
%
+% \begin{macro}{\@@_case_match:nn}
+% \begin{macro}[EXP]{\@@_case_match_aux:nn}
+% The code would get badly messed up if the number of items in |#1|
+% were not even, so we catch this case, then follow the same code as
+% \cs{regex_match:nnTF} but using \cs{@@_case_build:n} and without
+% returning a result.
+% \begin{macrocode}
+\cs_new_protected:Npn \@@_case_match:nn #1#2
+ {
+ \int_if_odd:nTF { \tl_count:n {#1} }
+ {
+ \__kernel_msg_error:nnxxxx { regex } { case-odd }
+ { \token_to_str:N \regex_case_match:nn } { code }
+ { \tl_count:n {#1} } { \tl_to_str:n {#1} }
+ }
+ {
+ \group_begin:
+ \@@_disable_submatches:
+ \@@_single_match:
+ \exp_args:Nx \@@_case_build:n
+ { \@@_tl_pairs_map:nNF {#1} \@@_case_match_aux:nn { } }
+ \int_gzero:N \g_@@_case_int
+ \@@_match:n {#2}
+ \group_end:
+ }
+ }
+\cs_new:Npn \@@_case_match_aux:nn #1#2 { \exp_not:n { {#1} } }
+% \end{macrocode}
+% \end{macro}
+% \end{macro}
+%
% \begin{macro}{\@@_count:nnN}
% Again, we don't care about submatches. Instead of aborting after the
% first \enquote{longest match} is found, we search for multiple
@@ -7617,8 +7824,15 @@
{ The~values~given~in~a~quantifier~must~be~in~order. }
% \end{macrocode}
%
-% Used when showing a regex.
+% Used in user commands, and when showing a regex.
% \begin{macrocode}
+\msg_new:nnnn { regex } { case-odd }
+ { #1~with~odd~number~of~items }
+ {
+ There~must~be~a~#2~part~for~each~regex:~
+ found~odd~number~of~items~(#3)~in\\
+ \iow_indent:n {#4}
+ }
\msg_new:nnn { regex } { show }
{
>~Compiled~regex~
diff --git a/l3kernel/testfiles/m3regex012.lvt b/l3kernel/testfiles/m3regex012.lvt
new file mode 100644
index 000000000..32f93aa47
--- /dev/null
+++ b/l3kernel/testfiles/m3regex012.lvt
@@ -0,0 +1,50 @@
+%
+% Copyright (C) 2021 The LaTeX Project
+\documentclass{minimal}
+\input{regression-test}
+
+\RequirePackage[enable-debug]{expl3}
+\ExplSyntaxOn
+\debug_on:n { check-declarations , deprecation , log-functions }
+\ExplSyntaxOff
+
+% \begin{document}
+
+\START
+\AUTHOR{Bruno Le Floch}
+\ExplSyntaxOn
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+
+\TEST { regex_case_match:nn }
+ {
+ \regex_set:Nn \l_tmpa_regex { [a-z]bc }
+ \cs_set_protected:Npn \test:n #1
+ {
+ \TYPE{#1:}
+ \regex_case_match:nnTF
+ {
+ \l_tmpa_regex { \TYPE{abc} }
+ { (?i) Y \w } { \TYPE{Y} }
+ { z \Z } { \TYPE{Z} }
+ }
+ {#1} { \TRUE } { \FALSE }
+ }
+ \test:n { }
+ \test:n { abc }
+ \test:n { Y abc }
+ \test:n { y abc }
+ \test:n { y bc }
+ \test:n { y ; bc }
+ \test:n { y ; bc z }
+ }
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+
+\TEST { regex_case_match:nn ~ error }
+ {
+ \regex_case_match:nn { * } { .. }
+ }
+
+%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
+\END
diff --git a/l3kernel/testfiles/m3clist006.tlg b/l3kernel/testfiles/m3regex012.tlg
similarity index 58%
copy from l3kernel/testfiles/m3clist006.tlg
copy to l3kernel/testfiles/m3regex012.tlg
index 7c17ac21a..d006b485b 100644
--- a/l3kernel/testfiles/m3clist006.tlg
+++ b/l3kernel/testfiles/m3regex012.tlg
@@ -2,26 +2,35 @@ This is a generated file for the LaTeX (2e + expl3) validation system.
Don't change this file in any respect.
Author: Bruno Le Floch
============================================================
-TEST 1: clist_reverse:n
+TEST 1: regex_case_match:nn
============================================================
-|| |{}|,|, |,{}| ,|{},|
-|,,, { },{ }, {},{}, |\c ,\a \b |
-============================================================
-============================================================
-TEST 2: clist_reverse:N
-============================================================
-Defining \c_test_clist on line ...
-\c ,{,},{ },\a \b
-TRUE
+:
FALSE
+abc:
+abc
TRUE
+Yabc:
+Y
TRUE
-============================================================
-============================================================
-TEST 3: clist_if_in:nnTF
-============================================================
-TRUE
+yabc:
+Y
TRUE
+ybc:
+abc
TRUE
+y;bc:
FALSE
+y;bcz:
+Z
+TRUE
+============================================================
+============================================================
+TEST 2: regex_case_match:nn error
+============================================================
+! LaTeX3 Error: \regex_case_match:nn with odd number of items
+For immediate help type H <return>.
+ ...
+l. ... }
+There must be a code part for each regex: found odd number of items (1) in
+ *
============================================================
More information about the latex3-commits
mailing list.