X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~mdw/git/disorder/blobdiff_plain/dd0f422a724b04d43e51be47c3517ce5fe4d99fe..6ebc4527c6a103d0532c08744fb916f951018413:/lib/macros.c

diff --git a/lib/macros.c b/lib/macros.c
index 2a5559c..e11c264 100644
--- a/lib/macros.c
+++ b/lib/macros.c
@@ -2,43 +2,37 @@
  * This file is part of DisOrder
  * Copyright (C) 2008 Richard Kettlewell
  *
- * This program is free software; you can redistribute it and/or modify
+ * This program is free software: you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
- * the Free Software Foundation; either version 2 of the License, or
+ * the Free Software Foundation, either version 3 of the License, or
  * (at your option) any later version.
- *
- * This program is distributed in the hope that it will be useful, but
- * WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * General Public License for more details.
- *
+ * 
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ * 
  * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
- * USA
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
  */
 
 /** @file lib/macros.c
  * @brief Macro expansion
  */
 
-#include <config.h>
-#include "types.h"
+#include "common.h"
 
-#include <string.h>
 #include <ctype.h>
-#include <assert.h>
-#include <stdio.h>
 #include <sys/stat.h>
 #include <fcntl.h>
 #include <unistd.h>
 #include <errno.h>
 
+#include "hash.h"
 #include "macros.h"
 #include "mem.h"
 #include "vector.h"
 #include "log.h"
-#include "hash.h"
 #include "sink.h"
 #include "syscalls.h"
 #include "printf.h"
@@ -107,6 +101,13 @@ static int mx__expand_macro(const struct expansion *e,
 
 /* Parsing ------------------------------------------------------------------ */
 
+static int next_non_whitespace(const char *input,
+                               const char *end) {
+  while(input < end && isspace((unsigned char)*input))
+    ++input;
+  return input < end ? *input : -1;
+}
+
 /** @brief Parse a template
  * @param filename Input filename (for diagnostics)
  * @param line Line number (use 1 on initial call)
@@ -126,18 +127,21 @@ const struct mx_node *mx_parse(const char *filename,
 			       int line,
 			       const char *input,
 			       const char *end) {
-  int braces, expansion_start_line, argument_start_line;
-  const char *argument_start, *argument_end, *p;
+  int braces, argument_start_line, obracket, cbracket;
+  const char *argument_start, *argument_end;
   struct mx_node_vector v[1];
   struct dynstr d[1];
   struct mx_node *head = 0, **tailp = &head, *e;
-  int omitted_terminator;
 
   if(!end)
     end = input + strlen(input);
   while(input < end) {
     if(*input != '@') {
-      expansion_start_line = line;
+      e = xmalloc(sizeof *e);
+      e->next = 0;
+      e->filename = filename;
+      e->line = line;
+      e->type = MX_TEXT;
       dynstr_init(d);
       /* Gather up text without any expansions in. */
       while(input < end && *input != '@') {
@@ -146,126 +150,119 @@ const struct mx_node *mx_parse(const char *filename,
 	dynstr_append(d, *input++);
       }
       dynstr_terminate(d);
-      e = xmalloc(sizeof *e);
-      e->next = 0;
-      e->filename = filename;
-      e->line = expansion_start_line;
-      e->type = MX_TEXT;
       e->text = d->vec;
       *tailp = e;
       tailp = &e->next;
       continue;
     }
-    mx_node_vector_init(v);
-    braces = 0;
-    p = input;
-    ++input;
-    expansion_start_line = line;
-    omitted_terminator = 0;
-    while(!omitted_terminator && input < end && *input != '@') {
-      /* Skip whitespace */
-      if(isspace((unsigned char)*input)) {
-	if(*input == '\n')
-	  ++line;
-	++input;
-	continue;
+    if(input + 1 < end)
+      switch(input[1]) {
+      case '@':
+        /* '@@' expands to '@' */
+        e = xmalloc(sizeof *e);
+        e->next = 0;
+        e->filename = filename;
+        e->line = line;
+        e->type = MX_TEXT;
+        e->text = "@";
+        *tailp = e;
+        tailp = &e->next;
+        input += 2;
+        continue;
+      case '#':
+        /* '@#' starts a (newline-eating comment), like dnl */
+        input += 2;
+        while(input < end && *input != '\n')
+          ++input;
+        if(*input == '\n') {
+          ++line;
+          ++input;
+        }
+        continue;
+      case '_':
+        /* '@_' expands to nothing.  It's there to allow dump to terminate
+         * expansions without having to know what follows. */
+        input += 2;
+        continue;
       }
-      if(*input == '{') {
-	/* This is a bracketed argument.  We'll walk over it counting
-	 * braces to figure out where the end is. */
-	++input;
-	argument_start = input;
-	argument_start_line = line;
-	while(input < end && (*input != '}' || braces > 0)) {
-	  switch(*input++) {
-	  case '{': ++braces; break;
-	  case '}': --braces; break;
-	  case '\n': ++line; break;
-	  }
-	}
-        /* If we run out of input without seeing a '}' that's an error */
-	if(input >= end)
-	  fatal(0, "%s:%d: unterminated expansion '%.*s'",
-		filename, argument_start_line,
-		(int)(input - argument_start), argument_start);
+    /* It's a full expansion */
+    ++input;
+    e = xmalloc(sizeof *e);
+    e->next = 0;
+    e->filename = filename;
+    e->line = line;
+    e->type = MX_EXPANSION;
+    /* Collect the expansion name.  Expansion names start with an alnum and
+     * consist of alnums and '-'.  We don't permit whitespace between the '@'
+     * and the name. */
+    dynstr_init(d);
+    if(input == end)
+      disorder_fatal(0, "%s:%d: invalid expansion syntax (truncated)",
+                     filename, e->line);
+    if(!isalnum((unsigned char)*input))
+      disorder_fatal(0, "%s:%d: invalid expansion syntax (unexpected %#x)",
+                     filename, e->line, (unsigned char)*input);
+    while(input < end && (isalnum((unsigned char)*input) || *input == '-'))
+      dynstr_append(d, *input++);
+    dynstr_terminate(d);
+    e->name = d->vec;
+    /* See what the bracket character is */
+    obracket = next_non_whitespace(input, end);
+    switch(obracket) {
+    case '(': cbracket = ')'; break;
+    case '[': cbracket = ']'; break;
+    case '{': cbracket = '}'; break;
+    default: cbracket = obracket = -1; break;      /* no arguments */
+    }
+    mx_node_vector_init(v);
+    if(obracket >= 0) {
+      /* Gather up arguments */
+      while(next_non_whitespace(input, end) == obracket) {
+        while(isspace((unsigned char)*input)) {
+          if(*input == '\n')
+            ++line;
+          ++input;
+        }
+        ++input;                        /* the bracket */
+        braces = 0;
+        /* Find the end of the argument */
+        argument_start = input;
+        argument_start_line = line;
+        while(input < end && (*input != cbracket || braces > 0)) {
+          const int c = *input++;
+
+          if(c == obracket)
+            ++braces;
+          else if(c == cbracket)
+            --braces;
+          else if(c == '\n')
+            ++line;
+        }
+        if(input >= end) {
+          /* We ran out of input without encountering a balanced cbracket */
+	  disorder_fatal(0, "%s:%d: unterminated expansion argument '%.*s'",
+                         filename, argument_start_line,
+                         (int)(input - argument_start), argument_start);
+        }
         /* Consistency check */
-	assert(*input == '}');
+        assert(*input == cbracket);
         /* Record the end of the argument */
         argument_end = input;
-        /* Step over the '}' */
+        /* Step over the cbracket */
 	++input;
-	if(input < end && isspace((unsigned char)*input)) {
-	  /* There is at least some whitespace after the '}'.  Look
-	   * ahead and see what is after all the whitespace. */
-	  for(p = input; p < end && isspace((unsigned char)*p); ++p)
-	    ;
-	  /* Now we are looking after the whitespace.  If it's
-	   * anything other than '{', including the end of the input,
-	   * then we infer that this expansion finished at the '}' we
-	   * just saw.  (NB that we don't move input forward to p -
-	   * the whitespace is NOT part of the expansion.) */
-	  if(p == end || *p != '{')
-	    omitted_terminator = 1;
-	}
-      } else {
-	/* We are looking at an unbracketed argument.  (A common example would
-	 * be the expansion or macro name.)  This is terminated by an '@'
-	 * (indicating the end of the expansion), a ':' (allowing a subsequent
-	 * unbracketed argument) or a '{' (allowing a bracketed argument).  The
-	 * end of the input will also do. */
-	argument_start = input;
-	argument_start_line = line;
-	while(input < end
-	      && *input != '@' && *input != '{' && *input != ':') {
-	  if(*input == '\n') ++line;
-	  ++input;
-	}
-        argument_end = input;
-        /* Trailing whitespace is not significant in unquoted arguments (and
-         * leading whitespace is eliminated by the whitespace skip above). */
-        while(argument_end > argument_start
-              && isspace((unsigned char)argument_end[-1]))
-          --argument_end;
-        /* Step over the ':' if that's what we see */
-	if(input < end && *input == ':')
-	  ++input;
+        /* Now we have an argument in [argument_start, argument_end), and we
+         * know its filename and initial line number.  This is sufficient to
+         * parse it. */
+        mx_node_vector_append(v, mx_parse(filename, argument_start_line,
+                                          argument_start, argument_end));
       }
-      /* Now we have an argument in [argument_start, argument_end), and we know
-       * its filename and initial line number.  This is sufficient to parse
-       * it. */
-      mx_node_vector_append(v, mx_parse(filename, argument_start_line,
-                                        argument_start, argument_end));
     }
-    /* We're at the end of an expansion.  We might have hit the end of the
-     * input, we might have hit an '@' or we might have matched the
-     * omitted_terminator criteria. */
-    if(input < end) {
-      if(!omitted_terminator) {
-        assert(*input == '@');
-        ++input;
-      }
-    }
-    /* @@ terminates this file */
-    if(v->nvec == 0)
-      break;
-    /* Currently we require that the first element, the expansion name, is
-     * always plain text.  Removing this restriction would raise some
-     * interesting possibilities but for the time being it is considered an
-     * error. */
-    if(v->vec[0]->type != MX_TEXT)
-      fatal(0, "%s:%d: expansion names may not themselves contain expansions",
-            v->vec[0]->filename, v->vec[0]->line);
     /* Guarantee a NULL terminator (for the case where there's more than one
      * argument) */
     mx_node_vector_terminate(v);
-    e = xmalloc(sizeof *e);
-    e->next = 0;
-    e->filename = filename;
-    e->line = expansion_start_line;
-    e->type = MX_EXPANSION;
-    e->name = v->vec[0]->text;
-    e->nargs = v->nvec - 1;
-    e->args = v->nvec > 1 ? &v->vec[1] : 0;
+    /* Fill in the remains of the node */
+    e->nargs = v->nvec;
+    e->args = v->vec;
     *tailp = e;
     tailp = &e->next;
   }
@@ -274,11 +271,14 @@ const struct mx_node *mx_parse(const char *filename,
 
 static void mx__dump(struct dynstr *d, const struct mx_node *m) {
   int n;
-  
+  const struct mx_node *mm;
+
   if(!m)
     return;
   switch(m->type) {
   case MX_TEXT:
+    if(m->text[0] == '@')
+      dynstr_append(d, '@');
     dynstr_append_string(d, m->text);
     break;
   case MX_EXPANSION:
@@ -289,7 +289,22 @@ static void mx__dump(struct dynstr *d, const struct mx_node *m) {
       mx__dump(d, m->args[n]);
       dynstr_append(d, '}');
     }
-    dynstr_append(d, '@');
+    /* If the next non-whitespace is '{', add @_ to stop it being
+     * misinterpreted */
+    mm = m->next;
+    while(mm && mm->type == MX_TEXT) {
+      switch(next_non_whitespace(mm->text, mm->text + strlen(mm->text))) {
+      case -1:
+        mm = mm->next;
+        continue;
+      case '{':
+        dynstr_append_string(d, "@_");
+        break;
+      default:
+        break;
+      }
+      break;
+    }
     break;
   default:
     assert(!"invalid m->type");
@@ -297,7 +312,10 @@ static void mx__dump(struct dynstr *d, const struct mx_node *m) {
   mx__dump(d, m->next);
 }
 
-/** @brief Dump a parse macro expansion to a string */
+/** @brief Dump a parse macro expansion to a string
+ *
+ * Not of production quality!  Only intended for testing!
+ */
 char *mx_dump(const struct mx_node *m) {
   struct dynstr d[1];
 
@@ -326,9 +344,7 @@ static int mx__register(unsigned flags,
   e->args = args;
   e->callback = callback;
   e->definition = definition;
-  return hash_add(expansions, name, &e,
-                  ((flags & EXP_TYPE_MASK) == EXP_MACRO)
-                      ? HASH_INSERT : HASH_INSERT_OR_REPLACE);
+  return hash_add(expansions, name, &e, HASH_INSERT_OR_REPLACE);
 }
 
 /** @brief Register a simple expansion rule
@@ -370,11 +386,13 @@ int mx_register_macro(const char *name,
                       const struct mx_node *definition) {
   if(mx__register(EXP_MACRO, name, nargs, nargs, args,  0/*callback*/,
                   definition)) {
+#if 0
     /* This locates the error to the definition, which may be a line or two
      * beyond the @define command itself.  The backtrace generated by
      * mx_expand() may help more. */
-    error(0, "%s:%d: duplicate definition of '%s'",
-          definition->filename, definition->line, name);
+    disorder_error(0, "%s:%d: duplicate definition of '%s'",
+                   definition->filename, definition->line, name);
+#endif
     return -2;
   }
   return 0;
@@ -414,17 +432,17 @@ int mx_expand(const struct mx_node *m,
   case MX_EXPANSION:
     rc = 0;
     if(!(e = hash_find(expansions, m->name))) {
-      error(0, "%s:%d: unknown expansion name '%s'",
-            m->filename, m->line, m->name);
+      disorder_error(0, "%s:%d: unknown expansion name '%s'",
+                     m->filename, m->line, m->name);
       if(sink_printf(output, "[['%s' unknown]]", m->name) < 0)
         return -1;
     } else if(m->nargs < e->min) {
-      error(0, "%s:%d: expansion '%s' requires %d args, only %d given",
-            m->filename, m->line, m->name, e->min, m->nargs);
+      disorder_error(0, "%s:%d: expansion '%s' requires %d args, only %d given",
+                     m->filename, m->line, m->name, e->min, m->nargs);
       if(sink_printf(output, "[['%s' too few args]]", m->name) < 0)
         return -1;
     } else if(m->nargs > e->max) {
-      error(0, "%s:%d: expansion '%s' takes at most %d args, but %d given",
+      disorder_error(0, "%s:%d: expansion '%s' takes at most %d args, but %d given",
             m->filename, m->line, m->name, e->max, m->nargs);
       if(sink_printf(output, "[['%s' too many args]]", m->name) < 0)
         return -1;
@@ -470,8 +488,8 @@ int mx_expand(const struct mx_node *m,
     if(rc) {
       /* For non-IO errors we generate some backtrace */
       if(rc != -1)
-        error(0,  "  ...in @%s at %s:%d",
-              m->name, m->filename, m->line);
+        disorder_error(0,  "  ...in @%s at %s:%d",
+                       m->name, m->filename, m->line);
       return rc;
     }
     break;
@@ -505,7 +523,7 @@ int mx_expandstr(const struct mx_node *m,
   } else
     *sp = 0;
   if(rc && rc != -1 && what)
-    error(0, "  ...in %s at %s:%d", what, m->filename, m->line);
+    disorder_error(0, "  ...in %s at %s:%d", what, m->filename, m->line);
   return rc;
 }
 
@@ -527,11 +545,11 @@ int mx_expand_file(const char *path,
   const struct mx_node *m;
 
   if((fd = open(path, O_RDONLY)) < 0)
-    fatal(errno, "error opening %s", path);
+    disorder_fatal(errno, "error opening %s", path);
   if(fstat(fd, &sb) < 0)
-    fatal(errno, "error statting %s", path);
+    disorder_fatal(errno, "error statting %s", path);
   if(!S_ISREG(sb.st_mode))
-    fatal(0, "%s: not a regular file", path);
+    disorder_fatal(0, "%s: not a regular file", path);
   sofar = 0;
   b = xmalloc_noptr(sb.st_size);
   while(sofar < sb.st_size) {
@@ -539,27 +557,59 @@ int mx_expand_file(const char *path,
     if(n > 0)
       sofar += n;
     else if(n == 0)
-      fatal(0, "unexpected EOF reading %s", path);
+      disorder_fatal(0, "unexpected EOF reading %s", path);
     else if(errno != EINTR)
-      fatal(errno, "error reading %s", path);
+      disorder_fatal(errno, "error reading %s", path);
   }
   xclose(fd);
   m = mx_parse(path, 1, b, b + sb.st_size);
   rc = mx_expand(m, output, u);
   if(rc && rc != -1)
     /* Mention inclusion in backtrace */
-    error(0, "  ...in inclusion of file '%s'", path);
+    disorder_error(0, "  ...in inclusion of file '%s'", path);
   return rc;
 }
 
-/** @brief Rewrite a parse tree substituting in macro arguments
+/* Macros ------------------------------------------------------------------- */
+
+/** @brief Rewrite a parse tree substituting sub-expansions
  * @param m Parse tree to rewrite (from macro definition)
+ * @param ... Name/value pairs to rewrite
+ * @return Rewritten parse tree
+ *
+ * The name/value pair list consists of pairs of strings and is terminated by
+ * (char *)0.  Names and values are both copied so need not survive the call.
+ */
+const struct mx_node *mx_rewritel(const struct mx_node *m,
+                                  ...) {
+  va_list ap;
+  hash *h = hash_new(sizeof (struct mx_node *));
+  const char *n, *v;
+  struct mx_node *e;
+
+  va_start(ap, m);
+  while((n = va_arg(ap, const char *))) {
+    v = va_arg(ap, const char *);
+    e = xmalloc(sizeof *e);
+    e->next = 0;
+    e->filename = m->filename;
+    e->line = m->line;
+    e->type = MX_TEXT;
+    e->text = xstrdup(v);
+    hash_add(h, n, &e, HASH_INSERT);
+    /* hash_add() copies n */
+  }
+  return mx_rewrite(m, h);
+}
+
+/** @brief Rewrite a parse tree substituting in macro arguments
+ * @param definition Parse tree to rewrite (from macro definition)
  * @param h Hash mapping argument names to argument values
  * @return Rewritten parse tree
  */
-static const struct mx_node *mx__rewrite(const struct mx_node *definition,
-                                         hash *h) {
-  const struct mx_node *head = 0, **tailp = &head, *argvalue, *m, *mm;
+const struct mx_node *mx_rewrite(const struct mx_node *definition,
+                                 hash *h) {
+  const struct mx_node *head = 0, **tailp = &head, *argvalue, *m, *mm, **ap;
   struct mx_node *nm;
   int n;
   
@@ -574,7 +624,7 @@ static const struct mx_node *mx__rewrite(const struct mx_node *definition,
       break;
     case MX_EXPANSION:
       if(m->nargs == 0
-         && (argvalue = *(const struct mx_node **)hash_find(h, m->name))) {
+         && (ap = hash_find(h, m->name))) {
         /* This expansion has no arguments and its name matches one of the
          * macro arguments.  (Even if it's a valid expansion name we override
          * it.)  We insert its value at this point.  We do NOT recursively
@@ -584,6 +634,7 @@ static const struct mx_node *mx__rewrite(const struct mx_node *definition,
          * We need to recreate the list structure but a shallow copy will
          * suffice here.
          */
+        argvalue = *ap;
         for(mm = argvalue; mm; mm = mm->next) {
           nm = xmalloc(sizeof *nm);
           *nm = *mm;
@@ -598,7 +649,7 @@ static const struct mx_node *mx__rewrite(const struct mx_node *definition,
         *nm = *m;
         nm->args = xcalloc(nm->nargs, sizeof (struct mx_node *));
         for(n = 0; n < nm->nargs; ++n)
-          nm->args[n] = mx__rewrite(m->args[n], h);
+          nm->args[n] = mx_rewrite(m->args[n], h);
         nm->next = 0;
         *tailp = nm;
         tailp = (const struct mx_node **)&nm->next;
@@ -632,7 +683,7 @@ static int mx__expand_macro(const struct expansion *e,
   for(n = 0; n < m->nargs; ++n)
     hash_add(h, e->args[n], &m->args[n], HASH_INSERT);
   /* Generate a rewritten parse tree */
-  m = mx__rewrite(e->definition, h);
+  m = mx_rewrite(e->definition, h);
   /* Expand the result */
   return mx_expand(m, output, u);
   /* mx_expand() will update the backtrace */