[6566] improve texindex backward compatibility with texinfo.tex

[Gavin D. Smith]

Revision: 6566
          http://svn.sv.gnu.org/viewvc/?view=rev&root=texinfo&revision=6566
Author:   gavin
Date:     2015-08-24 15:10:34 +0000 (Mon, 24 Aug 2015)
Log Message:
-----------
improve texindex backward compatibility with texinfo.tex

Modified Paths:
--------------
    trunk/ChangeLog
    trunk/doc/texinfo.tex
    trunk/texindex/texindex.awk
    trunk/texindex/ti.twjr

Modified: trunk/ChangeLog
===================================================================
--- trunk/ChangeLog     2015-08-24 11:58:56 UTC (rev 6565)
+++ trunk/ChangeLog     2015-08-24 15:10:34 UTC (rev 6566)
@@ -1,5 +1,20 @@
 2015-08-24  Gavin Smith  <address@hidden>
 
+       * ti/ti.twjr (extract_initial): Read initials delimited by 
+       balanced curly braces, to be compatible with older versions
+       of texinfo.tex.
+       (Special_chars): Remove backslash.
+       (Problem reported by Norbert Preining, bug-texinfo, on 6 July.)
+
+       * doc/texinfo.tex (\dosubindwrite): Output \ as 
+       {\indexbackslash}, not {\tt \indexbackslash}.
+       (\usemathbackslash): Removed.
+       (\indexnofonts): Likewise, output { and } as braced sequences, 
+       to be recognized by older versions of texindex.
+       (\printindex): Give these braced sequences meaning.
+
+2015-08-24  Gavin Smith  <address@hidden>
+
        * doc/texinfo-tex-test.texi: Add more examples of short indices, 
        and show some index initial glyphs.
 

Modified: trunk/doc/texinfo.tex
===================================================================
--- trunk/doc/texinfo.tex       2015-08-24 11:58:56 UTC (rev 6565)
+++ trunk/doc/texinfo.tex       2015-08-24 15:10:34 UTC (rev 6566)
@@ -346,10 +346,10 @@
     \normalturnoffactive  % \ in index entries must not stay \, e.g., if
                % the page break happens to be in the middle of an example.
                % We don't want .vr (or whatever) entries like this:
-               % \entry{{\tt \indexbackslash }acronym}{32}{\code {\acronym}}
+               % \entry{{\indexbackslash }acronym}{32}{\code {\acronym}}
                % "\acronym" won't work when it's read back in;
                % it needs to be
-               % {\code {{\tt \backslashcurfont }acronym}
+               % {\code {{\backslashcurfont }acronym}
     \shipout\vbox{%
       % Do this early so pdf references go to the beginning of the page.
       \ifpdfmakepagedest \pdfdest name{\the\pageno} xyz\fi
@@ -4450,7 +4450,8 @@
 \def\docodeindex#1{\edef\indexname{#1}\parsearg\singlecodeindexer}
 \def\singlecodeindexer #1{\doind{\indexname}{\code{#1}}}
 
-% Take care of Texinfo commands that can appear in an index entry.
+% Used when writing an index entry out to an index file, to prevent
+% expansion of Texinfo commands that can appear in an index entry.
 % Since there are some commands we want to expand, and others we don't,
 % we have to laboriously prevent expansion for those that we don't.
 %
@@ -4679,6 +4680,9 @@
 % For testing: output @{ and @} in index sort strings as \{ and \}.
 \newif\ifusebracesinindexes
 
+\let\indexlbrace\relax
+\let\indexrbrace\relax
+
 % \indexnofonts is used when outputting the strings to sort the index
 % by, and when constructing control sequence names.  It eliminates all
 % control sequences and just writes whatever the best ASCII sort string
@@ -4704,16 +4708,10 @@
   \def\_{\normalunderscore}%
   \def\-{}% @- shouldn't affect sorting
   %
-  % Unfortunately, texindex is not prepared to handle braces in the
-  % content at all.  So for index sorting, we map @{ and @} to strings
-  % starting with |, since that ASCII character is between ASCII { and }.
-  \ifusebracesinindexes
-    \def\lbracechar{\lbracecmd}%
-    \def\rbracechar{\rbracecmd}%
-  \else
-    \def\lbracechar{|a}%
-    \def\rbracechar{|b}%
-  \fi
+  % Unfortunately, texindex < 6.0 is not prepared to handle braces in the
+  % content at all, so these won't be sorted in ASCII order.
+  \def\lbracechar{{\indexlbrace}}%
+  \def\rbracechar{{\indexrbrace}}%
   \let\{=\lbracechar
   \let\}=\rbracechar
   %
@@ -4795,7 +4793,6 @@
 {\catcode`\`=\active
  \gdef\indexlquoteignore{\let`=\empty}}
 
-\let\indexbackslash=0  %overridden during \printindex.
 \let\SETmarginindex=\relax % put index entries in margin (undocumented)?
 
 % Most index entries go through here, but \dosubind is the general case.
@@ -4825,6 +4822,13 @@
   \fi
 }
 
+% Output \ as {\indexbackslash}, because \ is an escape character in
+% the index files.
+\let\indexbackslash=\relax
address@hidden \catcode`\\=\active
+  @address@hidden@address@hidden
+}
+
 % Write the entry in \toks0 to the index file:
 %
 \def\dosubindwrite{%
@@ -4835,8 +4839,8 @@
   %
   % Remember, we are within a group.
   \indexdummies % Must do this here, since \bf, etc expand at this stage
-  \def\backslashcurfont{\indexbackslash}% \indexbackslash isn't defined now
-      % so it will be output as is; and it will print as backslash.
+  \useindexbackslash % \indexbackslash isn't defined now so it will be output 
+                     % as is; and it will print as backslash.
   %
   % Process the index entry with all font commands turned off, to
   % get the string to sort by.
@@ -5000,7 +5004,9 @@
       % Index files are almost Texinfo source, but we use \ as the escape
       % character.  It would be better to use @, but that's too big a change
       % to make right now.
-      \def\indexbackslash{\backslashcurfont}%
+      \def\indexbackslash{\ttbackslash}%
+      \let\indexlbrace\{   % Likewise, set these sequences for braces
+      \let\indexrbrace\}   % used in the sort key.
       \catcode`\\ = 0
       \escapechar = `\\
       \begindoublecolumns
@@ -5014,9 +5020,6 @@
 % These macros are used by the sorted index file itself.
 % Change them to control the appearance of the index.
 
address@hidden \catcode`\\=13
address@hidden@address@hidden@address@hidden@backslash}}}}
-
 \let\normalhyphen=-
 {\catcode`\/=13 \catcode`\-=13 \catcode`\^=13 \catcode`\~=13 \catcode`\_=13
 \catcode`\|=13 \catcode`\<=13 \catcode`\>=13 \catcode`\+=13 \catcode`\"=13
@@ -5025,7 +5028,8 @@
   % Some changes for non-alphabetic characters.  Using the glyphs from the
   % math fonts looks more consistent than the typewriter font used elsewhere
   % for these characters.
-  \usemathbackslash
+  \def\indexbackslash{\math{\backslash}}%
+  \let\\=\indexbackslash
   %
   % Can't get bold backslash so don't use bold forward slash
   \catcode`\/=13

Modified: trunk/texindex/texindex.awk
===================================================================
--- trunk/texindex/texindex.awk 2015-08-24 11:58:56 UTC (rev 6565)
+++ trunk/texindex/texindex.awk 2015-08-24 15:10:34 UTC (rev 6566)
@@ -161,7 +161,7 @@
   EXIT_SUCCESS = 0
   EXIT_FAILURE = 1
   
-  Texindex_version = "5.9.93"
+  Texindex_version = "6.0dev"
   if (! Invocation_name) {
     # provide fallback in case it's not passed in.
     Invocation_name = "texindex"
@@ -268,18 +268,30 @@
 function extract_initial(key,  initial, nextchar, i, l, kchars)
 {
   l = char_split(key, kchars)
-  for (i = 1; i <= l; i++)
-    if (kchars[i] != "{")
-      break
+  if (l >= 3 && kchars[2] == "{") {
+    bracecount = 1
+    i = 3
+    while (bracecount > 0 && i <= l) {
+      if (kchars[i] == "{")
+        bracecount++
+      else if (kchars[i] == "}")
+        bracecount--
+      i++
+    }
+    if (i > l)
+      fatal(_"%s:%d: Bad key %s in record\n", FILENAME, FNR, key)
+    initial = substr(key, 2, i - 2)
+  } else if (kchars[2] == Command_char) {
+    nextchar = kchars[3]
+    if (initial == Command_char && (nextchar == "{" || nextchar == "}"))
+      initial = substr(key, 2, 3)
+    else {
+      initial = toupper(nextchar)
+    }
+  } else {
+    initial = toupper(kchars[2])
+  }
 
-  if (i > l)
-    fatal(_"%s:%d: Bad key %s in record\n", FILENAME, FNR, key)
-
-  initial = toupper(kchars[i])
-  nextchar = kchars[i+1]
-  if (initial == Command_char && (nextchar == "{" || nextchar == "}"))
-    initial = nextchar
-
   return initial
 }
 function field_split( \

Modified: trunk/texindex/ti.twjr
===================================================================
--- trunk/texindex/ti.twjr      2015-08-24 11:58:56 UTC (rev 6565)
+++ trunk/texindex/ti.twjr      2015-08-24 15:10:34 UTC (rev 6566)
@@ -411,7 +411,7 @@
 @samp{entry}.
 
 @cindex @code{Special_chars} variable
address@hidden are the three characters that must be preceded by
address@hidden are the two characters that must be preceded by
 the command character inside the first key.
 
 @cindex @code{beginfile()} function
@@ -443,7 +443,7 @@
       || substr($0, 2, 5) != "entry")
     fatal(_"%s is not a Texinfo index file\n", filename)
 
-  Special_chars = "{}" Command_char
+  Special_chars = "{}"
 }
 @
 
@@ -512,12 +512,14 @@
 initial = extract_initial($0)
 @
 
-Extracting the initial is mildly complicated.  Braces can be nested, and
-in particular the very first field of the sort key can be an open brace.
-So it is necessary to skip leading open braces until we encounter the
-first real character.  This in turn could be @address@hidden or @address@hidden
-preceded by the command character, or another character.
+The key argument here is the rest of the line after @samp{\entry}, 
+starting with an open brace.
 
+The very first field character of the sort key can be an open brace.
+If so, we extract the component of the sort key surrounded by balanced 
+braces.  We don't account for address@hidden or address@hidden inside this 
component, as 
address@hidden isn't expected to produce such output.
+
 An example can be seen in what older versions of @file{texinfo.tex}
 generated if you needed to index a real backslash, namely an input line
 something like the following:
@@ -526,8 +528,10 @@
 address@hidden@{\tt \indexbackslash @} 
(backslash)@address@hidden@address@hidden @address@hidden @dots{}
 @end example
 
-Fortunately, the first non-brace character is a backslash, and that is
-also the correct initial.
+Earlier versions of @command{texindex} took the the first non-brace 
+character as the initial, in this example @samp{\}, and output it as 
address@hidden; this was not, however, a control sequence recognized by the 
+older versions of @file{texinfo.tex}.
 
 @cindex @code{extract_initial()} function
 @cindex @code{char_split()} function
@@ -535,18 +539,30 @@
 function extract_initial(key,  initial, nextchar, i, l, kchars)
 {
   l = char_split(key, kchars)
-  for (i = 1; i <= l; i++)
-    if (kchars[i] != "{")
-      break
+  if (l >= 3 && kchars[2] == "{") {
+    bracecount = 1
+    i = 3
+    while (bracecount > 0 && i <= l) {
+      if (kchars[i] == "{")
+        bracecount++
+      else if (kchars[i] == "}")
+        bracecount--
+      i++
+    }
+    if (i > l)
+      fatal(_"%s:%d: Bad key %s in record\n", FILENAME, FNR, key)
+    initial = substr(key, 2, i - 2)
+  } else if (kchars[2] == Command_char) {
+    nextchar = kchars[3]
+    if (initial == Command_char && (nextchar == "{" || nextchar == "}"))
+      initial = substr(key, 2, 3)
+    else {
+      initial = toupper(nextchar)
+    }
+  } else {
+    initial = toupper(kchars[2])
+  }
 
-  if (i > l)
-    fatal(_"%s:%d: Bad key %s in record\n", FILENAME, FNR, key)
-
-  initial = toupper(kchars[i])
-  nextchar = kchars[i+1]
-  if (initial == Command_char && (nextchar == "{" || nextchar == "}"))
-    initial = nextchar
-
   return initial
 }
 @