[latex3-commits] [git/LaTeX3-latex3-latex3] master: Extend final-sigma support to 8-bit engines (937bcb6af)

[Joseph Wright]

Repository : https://github.com/latex3/latex3
On branch  : master
Link       : https://github.com/latex3/latex3/commit/937bcb6af4cfd0212cfcbc2e25ad141730ae07cb

>---------------------------------------------------------------

commit 937bcb6af4cfd0212cfcbc2e25ad141730ae07cb
Author: Joseph Wright <joseph.wright at morningstar2.co.uk>
Date:   Thu Oct 15 13:51:00 2020 +0100

    Extend final-sigma support to 8-bit engines


>---------------------------------------------------------------

937bcb6af4cfd0212cfcbc2e25ad141730ae07cb
 l3kernel/l3text-case.dtx | 77 ++++++++++++++++++++++++++++++++++++++----------
 1 file changed, 62 insertions(+), 15 deletions(-)

diff --git a/l3kernel/l3text-case.dtx b/l3kernel/l3text-case.dtx
index febd34ccd..a0840cc85 100644
--- a/l3kernel/l3text-case.dtx
+++ b/l3kernel/l3text-case.dtx
@@ -155,6 +155,9 @@
 % \begin{macro}[EXP]{\@@_change_case_lower_sigma:nnnN}
 % \begin{macro}[EXP]{\@@_change_case_lower_sigma:nnNw}
 % \begin{macro}[EXP]{\@@_change_case_lower_sigma:NnnN}
+% \begin{macro}[EXP]{\@@_change_case_lower_sigma:nnnNN}
+% \begin{macro}[EXP]{\@@_change_case_lower_sigma:nnw}
+% \begin{macro}[EXP]{\@@_change_case_lower_sigma:nnN}
 % \begin{macro}[EXP]
 %   {\@@_change_case_char_title:nnN, \@@_change_case_char_titleonly:nnN}
 % \begin{macro}[EXP]
@@ -461,16 +464,10 @@
 %   is there the special case of a terminal sigma. If not, then we pass to
 %   a simple character mapping.
 %    \begin{macrocode}
-\cs_new:Npx \@@_change_case_char_lower:nnN #1#2#3
+\cs_new:Npn \@@_change_case_char_lower:nnN #1#2#3
   {
-    \exp_not:N \cs_if_exist_use:cF { @@_change_case_lower_ #2 :nnnN }
-      {
-        \bool_lazy_or:nnTF
-          { \sys_if_engine_luatex_p: }
-          { \sys_if_engine_xetex_p: }
-          { \exp_not:N \@@_change_case_lower_sigma:nnnN }
-          { \exp_not:N \@@_change_case_char:nnnN }
-       }
+    \cs_if_exist_use:cF { @@_change_case_lower_ #2 :nnnN }
+      { \@@_change_case_lower_sigma:nnnN }
         {#1} {#1} {#2} #3
   }
 \cs_new:Npn \@@_change_case_char_upper:nnN #1#2#3
@@ -485,7 +482,7 @@
 %   then there is a look-ahead phase: the logic here is simply based on letters.
 %   The one exception is Dutch: see below.
 %    \begin{macrocode}
-\bool_lazy_or:nnT
+\bool_lazy_or:nnTF
   { \sys_if_engine_luatex_p: }
   { \sys_if_engine_xetex_p: }
   {
@@ -518,6 +515,48 @@
       }
   }
 %    \end{macrocode}
+%   In the $8$-bit engines, we have to look ahead once we find the first byte of the 
+%   possible hit.
+%    \begin{macrocode}
+  {
+    \cs_new:Npn \@@_change_case_lower_sigma:nnnN #1#2#3#4
+      {
+        \int_compare:nNnTF { `#4 } = { "CE }
+          { \@@_change_case_lower_sigma:nnnNN  }
+          { \@@_change_case_char:nnnN }
+            {#1} {#2} {#3} #4
+      }
+    \cs_new:Npn \@@_change_case_lower_sigma:nnnNN #1#2#3#4#5
+      {
+        \int_compare:nNnTF { `#5 } = { "A3 }
+          { \@@_change_case_lower_sigma:nnw {#2} {#3}  }
+          { \@@_change_case_char:nnnN {#1} {#2} {#3} #4#5 }
+      }
+    \cs_new:Npn \@@_change_case_lower_sigma:nnw #1#2#3 \q_@@_recursion_stop
+      {
+        \tl_if_head_is_N_type:nTF {#3}
+          { \@@_change_case_lower_sigma:nnN }
+          {
+            \@@_change_case_store:V \c_@@_final_sigma_tl
+            \@@_change_case_loop:nnw
+          }
+        {#1} {#2} #3 \q_@@_recursion_stop
+      }
+    \cs_new:Npn \@@_change_case_lower_sigma:nnN #1#2#3
+      {
+        \bool_lazy_or:nnTF
+          { \token_if_letter_p:N #3 }
+          {
+            \bool_lazy_and_p:nn
+              { \token_if_active_p:N #3 }
+              { \int_compare_p:nNn { `#3 } > { "80 } }
+          }
+          { \@@_change_case_store:V \c_@@_sigma_tl }
+          { \@@_change_case_store:V \c_@@_final_sigma_tl }
+        \@@_change_case_loop:nnw {#1} {#2} #3
+      }
+  }
+%    \end{macrocode}
 %   For titlecasing, we need to fully expand the new character to see if it
 %   is a letter (or active) But that means looking ahead in the $8$-bit case, so
 %   we have to grab the required tokens up-front. Life is a lot easier for Unicode
@@ -664,6 +703,9 @@
 % \end{macro}
 % \end{macro}
 % \end{macro}
+% \end{macro}
+% \end{macro}
+% \end{macro}
 %
 % \begin{macro}[EXP]{\@@_change_case_upper_de-alt:nnnN}
 % \begin{macro}[EXP]{\@@_change_case_upper_de-alt:nnnNN}
@@ -1198,14 +1240,17 @@
 %
 % \begin{variable}
 %   {
-%     \c_@@_dotless_i_tl ,
-%     \c_@@_dotted_I_tl  ,
-%     \c_@@_i_ogonek_tl  ,
-%     \c_@@_I_ogonek_tl  ,
+%     \c_@@_dotless_i_tl   ,
+%     \c_@@_dotted_I_tl    ,
+%     \c_@@_i_ogonek_tl    ,
+%     \c_@@_I_ogonek_tl    ,
+%     \c_@@_final_sigma_tl ,
+%     \c_@@_sigma_tl       ,
 %     \c_@@_grosses_Eszett_tl
 %   }
 %  For cases where there is an $8$-bit option in the |T1| font set up,
-%  a variant is provided in both cases.
+%  a variant is provided in both cases. There are also a few extras for
+%  |LGR|.
 %    \begin{macrocode}
 \group_begin:
   \bool_lazy_or:nnF
@@ -1238,6 +1283,8 @@
       \@@_tmp:w \c_@@_dotted_I_tl       { 0130 }
       \@@_tmp:w \c_@@_i_ogonek_tl       { 012F }
       \@@_tmp:w \c_@@_I_ogonek_tl       { 012E }
+      \@@_tmp:w \c_@@_final_sigma_tl    { 03C2 }
+      \@@_tmp:w \c_@@_sigma_tl          { 03C3 }
       \@@_tmp:w \c_@@_grosses_Eszett_tl { 1E9E }
     }
 \group_end: