[6121] parsetexi register_command_arg

[Gavin D. Smith]

Revision: 6121
          http://svn.sv.gnu.org/viewvc/?view=rev&root=texinfo&revision=6121
Author:   gavin
Date:     2015-02-12 22:35:56 +0000 (Thu, 12 Feb 2015)
Log Message:
-----------
parsetexi register_command_arg

Modified Paths:
--------------
    trunk/parsetexi/ChangeLog
    trunk/parsetexi/dump_perl.c
    trunk/parsetexi/end_line.c
    trunk/parsetexi/extra.c
    trunk/parsetexi/makeinfo-debug.txt
    trunk/parsetexi/parser.c
    trunk/parsetexi/parser.h
    trunk/parsetexi/separator.c
    trunk/parsetexi/tree.c
    trunk/parsetexi/tree.h
    trunk/parsetexi/tree_types.h

Added Paths:
-----------
    trunk/parsetexi/test-files/email.texi

Modified: trunk/parsetexi/ChangeLog
===================================================================
--- trunk/parsetexi/ChangeLog   2015-02-12 19:36:41 UTC (rev 6120)
+++ trunk/parsetexi/ChangeLog   2015-02-12 22:35:56 UTC (rev 6121)
@@ -1,5 +1,12 @@
 2015-02-12  Gavin Smith  <address@hidden>
 
+       * separator.c (register_command_arg): New function.
+       (handle_close_brace, handle_comma) <brace commands with positive 
+       number of args>: Call it.
+       * extra.c (add_extra_key_contents_array): New function.
+
+2015-02-12  Gavin Smith  <address@hidden>
+
        * command_data.txt <brace commands>: Add arg types for all 
        commands (some of these could change).
 

Modified: trunk/parsetexi/dump_perl.c
===================================================================
--- trunk/parsetexi/dump_perl.c 2015-02-12 19:36:41 UTC (rev 6120)
+++ trunk/parsetexi/dump_perl.c 2015-02-12 22:35:56 UTC (rev 6121)
@@ -69,13 +69,23 @@
 dump_args (ELEMENT *e, TEXT *text)
 {
   int i;
+  int not_in_tree = 0;
+
+  /* Don't set routing information if not dumping from a route
+     directly from the root, that is via an extra value. */
+  if (e->parent_type == route_not_in_tree
+      || (e->parent_type == route_uninitialized && e->parent))
+    not_in_tree = 1;
   text_append_n (text, "[\n", 2);
   indent += 2;
 
   for (i = 0; i < e->args.number; i++)
     {
-      e->args.list[i]->parent_type = route_args;
-      e->args.list[i]->index_in_parent = i;
+      if (!not_in_tree)
+        {
+          e->args.list[i]->parent_type = route_args;
+          e->args.list[i]->index_in_parent = i;
+        }
 
       dump_indent (text);
       dump_element (e->args.list[i], text);
@@ -91,13 +101,24 @@
 dump_contents (ELEMENT *e, TEXT *text)
 {
   int i;
+  int not_in_tree = 0;
+
+  /* Don't set routing information if not dumping from a route
+     directly from the root, that is via an extra value. */
+  if (e->parent_type == route_not_in_tree
+      || (e->parent_type == route_uninitialized && e->parent))
+    not_in_tree = 1;
+
   text_append_n (text, "[\n", 2);
   indent += 2;
 
   for (i = 0; i < e->contents.number; i++)
     {
-      e->contents.list[i]->parent_type = route_contents;
-      e->contents.list[i]->index_in_parent = i;
+      if (!not_in_tree)
+        {
+          e->contents.list[i]->parent_type = route_contents;
+          e->contents.list[i]->index_in_parent = i;
+        }
 
       dump_indent (text);
       dump_element (e->contents.list[i], text);
@@ -289,6 +310,7 @@
             {
               switch (e->extra[i].type)
                 {
+                  int j;
                 case extra_element:
                   dump_element (e->extra[i].value, text);
                   break;
@@ -298,6 +320,22 @@
                   text_append (text, "' => ");
                   dump_contents (e->extra[i].value, text);
                   break;
+                case extra_element_contents_array:
+                  /* Like extra_element_contents, but this time output an array
+                     of arrays (instead of an array). */
+                  text_append_n (text, "'", 1);
+                  text_append (text, e->extra[i].key);
+                  text_append (text, "' => [\n");
+                  indent += 2;
+                  for (j = 0; j < e->extra[i].value->contents.number; j++)
+                    {
+                      dump_indent (text);
+                      dump_contents (e->extra[i].value->contents.list[j], 
+                                     text);
+                    }
+                  indent -= 2;
+                  text_append (text, "],");
+                  break;
                 default:
                   abort ();
                 }

Modified: trunk/parsetexi/end_line.c
===================================================================
--- trunk/parsetexi/end_line.c  2015-02-12 19:36:41 UTC (rev 6120)
+++ trunk/parsetexi/end_line.c  2015-02-12 22:35:56 UTC (rev 6121)
@@ -27,37 +27,6 @@
 #include "indices.h"
 #include "errors.h"
 
-// 5467, also in Common.pm 1334
-// TODO: Check the behaviour here is the same
-/* Return a new element whose contents are the same as those of ORIGINAL,
-   but with some elements representing empty spaces removed.  Elements like 
-   these are used to represent some of the "content" extra keys. */
-ELEMENT *
-trim_spaces_comment_from_content (ELEMENT *original)
-{
-  ELEMENT *trimmed;
-  int i;
-
-  trimmed = new_element (ET_NONE);
-  trimmed->parent_type = route_not_in_tree;
-  for (i = 0; i < original->contents.number; i++)
-    {
-      if (original->contents.list[i]->type
-          != ET_empty_spaces_after_command
-          && original->contents.list[i]->type != ET_spaces_at_end)
-        {
-          /* FIXME: Is this safe to serialize? */
-          /* For example, if there are extra keys in the elements under each 
-             argument?  They may not be set in a copy.
-             Hopefully there aren't many extra keys set on commands in 
-             node names. */
-          add_to_element_contents (trimmed, original->contents.list[i]);
-        }
-    }
-
-  return trimmed;
-}
-
 static int
 is_decimal_number (char *string)
 {

Modified: trunk/parsetexi/extra.c
===================================================================
--- trunk/parsetexi/extra.c     2015-02-12 19:36:41 UTC (rev 6120)
+++ trunk/parsetexi/extra.c     2015-02-12 22:35:56 UTC (rev 6121)
@@ -51,6 +51,13 @@
   e->extra[e->extra_number - 1].type = extra_element_contents;
 }
 
+void
+add_extra_key_contents_array (ELEMENT *e, char *key, ELEMENT *value)
+{
+  add_extra_key (e, key, value);
+  e->extra[e->extra_number - 1].type = extra_element_contents_array;
+}
+
 /* Add an extra key that is a reference to the text field of another
    element. */
 void

Modified: trunk/parsetexi/makeinfo-debug.txt
===================================================================
--- trunk/parsetexi/makeinfo-debug.txt  2015-02-12 19:36:41 UTC (rev 6120)
+++ trunk/parsetexi/makeinfo-debug.txt  2015-02-12 22:35:56 UTC (rev 6121)
@@ -3,6 +3,7 @@
 $makeinfo -c DUMP_TREE=1 multitparseable.texi 2>&1 |less
 
 makeinfo -c TEXINFO_OUTPUT_FORMAT=debugtree mltitble.texi
+----- often more readable than DUMP_TREE, but no "extra"
 
 makeinfo -c DEBUG=1 -c TEXINFO_OUTPUT_FORMAT=parse mltitble.texi
 

Modified: trunk/parsetexi/parser.c
===================================================================
--- trunk/parsetexi/parser.c    2015-02-12 19:36:41 UTC (rev 6120)
+++ trunk/parsetexi/parser.c    2015-02-12 22:35:56 UTC (rev 6121)
@@ -380,10 +380,15 @@
 }
 
 /* 2149 */
-/* Split any trailing whitespace on the last element in a line into its
-   own element, ET_spaces_at_end by default.  This helps with line
-   argument parsing as there will be no leading or trailing spaces.
+/* Split any trailing whitespace on the last contents child of CURRENT into
+   own element, ET_spaces_at_end by default.
+  
+   This is used for the argument to a line command, and for the arguments to a 
+   brace command taking a given number of arguments.
 
+   This helps with argument parsing as there will be no leading or trailing 
+   spaces.
+
    Also, "to help expansion disregard unuseful spaces".  Could that mean
    macro expansion? */
 void
@@ -458,6 +463,39 @@
     }
 }
 
+// 5467, also in Common.pm 1334
+// TODO: Check the behaviour here is the same
+/* Return a new element whose contents are the same as those of ORIGINAL,
+   but with some elements representing empty spaces removed.  Elements like 
+   these are used to represent some of the "content" extra keys. */
+ELEMENT *
+trim_spaces_comment_from_content (ELEMENT *original)
+{
+  ELEMENT *trimmed;
+  int i;
+
+  trimmed = new_element (ET_NONE);
+  trimmed->parent_type = route_not_in_tree;
+  for (i = 0; i < original->contents.number; i++)
+    {
+      if (original->contents.list[i]->type != ET_empty_spaces_after_command
+        && original->contents.list[i]->type != ET_spaces_at_end
+        && original->contents.list[i]->type != ET_empty_spaces_before_argument)
+        {
+          /* FIXME: Is this safe to serialize? */
+          /* For example, if there are extra keys in the elements under each 
+             argument?  They may not be set in a copy.
+             Hopefully there aren't many extra keys set on commands in 
+             node names. */
+          //add_to_element_contents (trimmed, original->contents.list[i]);
+          add_to_contents_as_array (trimmed, original->contents.list[i]);
+        }
+    }
+
+  return trimmed;
+}
+
+
 /* 3491 */
 /* Add an "ET_empty_line_after_command" element containing the whitespace at 
    the beginning of the rest of the line.  This element can be later changed 
to 

Modified: trunk/parsetexi/parser.h
===================================================================
--- trunk/parsetexi/parser.h    2015-02-12 19:36:41 UTC (rev 6120)
+++ trunk/parsetexi/parser.h    2015-02-12 22:35:56 UTC (rev 6121)
@@ -42,6 +42,7 @@
 char *read_command_name (char **ptr);
 ELEMENT *merge_text (ELEMENT *current, char *text);
 void start_empty_line_after_command (ELEMENT *current, char **line_inout);
+ELEMENT *trim_spaces_comment_from_content (ELEMENT *original);
 
 extern const char *whitespace_chars, *whitespace_chars_except_newline;
 extern const char *digit_chars;
@@ -65,6 +66,7 @@
 /* In extra.c */
 void add_extra_key_element (ELEMENT *e, char *key, ELEMENT *value);
 void add_extra_key_contents (ELEMENT *e, char *key, ELEMENT *value);
+void add_extra_key_contents_array (ELEMENT *e, char *key, ELEMENT *value);
 void add_extra_key_text (ELEMENT *e, char *key, ELEMENT *value);
 void add_extra_key_index_entry (ELEMENT *e, char *key, INDEX_ENTRY_REF *value);
 void add_extra_key_misc_args (ELEMENT *e, char *key, ELEMENT *value);

Modified: trunk/parsetexi/separator.c
===================================================================
--- trunk/parsetexi/separator.c 2015-02-12 19:36:41 UTC (rev 6120)
+++ trunk/parsetexi/separator.c 2015-02-12 22:35:56 UTC (rev 6121)
@@ -21,6 +21,39 @@
 #include "tree.h"
 #include "text.h"
 
+// 3600
+/* Add the contents of CURRENT as an element to the extra value with
+   key KEY, except that some "empty space" elements are removed.  Used for
+   'brace_command_contents' for the arguments to a brace command, and
+   'block_command_line_contents' for the arguments to a block line command.
+
+   For a brace command $element, $element->{'args'} has pretty much the same 
+   information as $element->{'extra'}->{'brace_command_contents'}. */
+
+static void
+register_command_arg (ELEMENT *current, char *key)
+{
+  ELEMENT *value;
+  ELEMENT *new;
+  KEY_PAIR *k;
+
+  /* FIXME: Could we add all the command args together, instead of one-by-one,
+     to avoid having to look for the extra value every time? */
+  k = lookup_extra_key (current->parent, key);
+  if (k)
+    value = k->value;
+  else
+    {
+      value = new_element (ET_NONE);
+      value->parent_type = route_not_in_tree;
+      add_extra_key_contents_array (current->parent, key, value);
+    }
+
+  new = trim_spaces_comment_from_content (current);
+
+  add_to_contents_as_array (value, new);
+}
+
 /* 4888 */
 ELEMENT *
 handle_open_brace (ELEMENT *current, char **line_inout)
@@ -29,8 +62,7 @@
 
   abort_empty_line (&current, NULL);
   /* 4890 */
-  if (command_flags(current) & CF_brace
-      /* || definfoenclose */ )
+  if (command_flags(current) & CF_brace)
     {
       enum command_id command;
       ELEMENT *arg;
@@ -86,9 +118,12 @@
       else /* not context brace */
         {
           current->type = ET_brace_command_arg;
-          /* If this command takes more than one argument (why??) */
-          /* TODO: Also "simple text commands" 425 */
-          if (command_data(command).data > 1)
+
+          /* Commands which are said to take a positive number of arguments
+             disregard leading and trailing whitespace.  In 
+             'handle_close_brace', the 'brace_command_contents' array
+             is set.  */
+          if (command_data(command).data > 0)
             {
               ELEMENT *e;
               e = new_element (ET_empty_spaces_before_argument);
@@ -96,9 +131,12 @@
               add_extra_key_element (current->parent,
                                      "spaces_before_argument", e);
             }
-          /* 4956 - TODO inline commands */
+          else if (command_data(command).data == BRACE_inline)
+            {
+              // 4956
+            }
         }
-
+      debug ("OPENED");
     }
 
   /* 4967 */
@@ -169,6 +207,14 @@
           /* The Perl code here checks that the popped context and the
              parent command match as strings. */
         }
+      else if (command_data(current->parent->cmd).data > 0)
+        {
+          // 5033
+          isolate_last_space (current, 0);
+          register_command_arg (current, "brace_command_contents");
+          //remove_empty_content_arguments ();
+        }
+
       closed_command = current->parent->cmd;
       debug ("CLOSING(brace) %s", command_data(closed_command).cmdname);
 
@@ -261,7 +307,15 @@
   abort_empty_line (&current, NULL);
 
   /* Register brace_command_contents or block_command_line_contents in extra 
-     key.  Hopefully we won't have to do this. */
+     key. */
+  if (command_flags(current->parent) & CF_brace
+      && command_data(current->parent->cmd).data > 0)
+    {
+      // 5033
+      isolate_last_space (current, 0);
+      register_command_arg (current, "brace_command_contents");
+      //remove_empty_content_arguments ();
+    }
 
   type = current->type;
   current = current->parent;

Added: trunk/parsetexi/test-files/email.texi
===================================================================
--- trunk/parsetexi/test-files/email.texi                               (rev 0)
+++ trunk/parsetexi/test-files/email.texi       2015-02-12 22:35:56 UTC (rev 
6121)
@@ -0,0 +1 @@
address@hidden, same place}

Modified: trunk/parsetexi/tree.c
===================================================================
--- trunk/parsetexi/tree.c      2015-02-12 19:36:41 UTC (rev 6120)
+++ trunk/parsetexi/tree.c      2015-02-12 22:35:56 UTC (rev 6121)
@@ -77,7 +77,18 @@
   e->parent = parent;
 }
 
+/* Special purpose function for when we are only using PARENT as an
+   array, and we don't want to overwrite E->parent. */
 void
+add_to_contents_as_array (ELEMENT *parent, ELEMENT *e)
+{
+  ELEMENT_LIST *list = &parent->contents;
+  reallocate_list (list);
+
+  list->list[list->number++] = e;
+}
+
+void
 add_to_element_args (ELEMENT *parent, ELEMENT *e)
 {
   ELEMENT_LIST *list = &parent->args;

Modified: trunk/parsetexi/tree.h
===================================================================
--- trunk/parsetexi/tree.h      2015-02-12 19:36:41 UTC (rev 6120)
+++ trunk/parsetexi/tree.h      2015-02-12 22:35:56 UTC (rev 6121)
@@ -16,6 +16,7 @@
 
 ELEMENT *new_element (enum element_type type);
 void add_to_element_contents (ELEMENT *parent, ELEMENT *e);
+void add_to_contents_as_array (ELEMENT *parent, ELEMENT *e);
 void add_to_element_args (ELEMENT *parent, ELEMENT *e);
 void insert_into_contents (ELEMENT *parent, ELEMENT *e, int where);
 ELEMENT *remove_from_contents (ELEMENT *parent, int where);

Modified: trunk/parsetexi/tree_types.h
===================================================================
--- trunk/parsetexi/tree_types.h        2015-02-12 19:36:41 UTC (rev 6120)
+++ trunk/parsetexi/tree_types.h        2015-02-12 22:35:56 UTC (rev 6121)
@@ -27,6 +27,7 @@
 enum extra_type {
     extra_element,
     extra_element_contents,
+    extra_element_contents_array,
     extra_element_text,
     extra_index_entry,
     extra_misc_args,