Man-page back end for Halibut. Also, a couple of additional markup
[sgt/halibut] / input.c
diff --git a/input.c b/input.c
index d4ecaeb..6cad0f7 100644 (file)
--- a/input.c
+++ b/input.c
@@ -180,11 +180,14 @@ enum {
     c_copyright,                      /* copyright statement */
     c_cw,                             /* weak code */
     c_date,                           /* document processing date */
+    c_dd,                             /* description list: description */
     c_define,                         /* macro definition */
+    c_dt,                             /* description list: described thing */
     c_e,                              /* emphasis */
     c_i,                              /* visible index mark */
     c_ii,                             /* uncapitalised visible index mark */
     c_k,                              /* uncapitalised cross-reference */
+    c_lcont,                          /* continuation para(s) for list item */
     c_n,                              /* numbered list */
     c_nocite,                         /* bibliography trickery */
     c_preamble,                               /* document preamble text */
@@ -245,11 +248,14 @@ static void match_kw(token *tok) {
        {"copyright", c_copyright},    /* copyright statement */
        {"cw", c_cw},                  /* weak code */
        {"date", c_date},              /* document processing date */
+       {"dd", c_dd},                  /* description list: description */
        {"define", c_define},          /* macro definition */
+       {"dt", c_dt},                  /* description list: described thing */
        {"e", c_e},                    /* emphasis */
        {"i", c_i},                    /* visible index mark */
        {"ii", c_ii},                  /* uncapitalised visible index mark */
        {"k", c_k},                    /* uncapitalised cross-reference */
+       {"lcont", c_lcont},            /* continuation para(s) for list item */
        {"n", c_n},                    /* numbered list */
        {"nocite", c_nocite},          /* bibliography trickery */
        {"preamble", c_preamble},      /* document preamble text */
@@ -506,6 +512,7 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
     int already;
     int iswhite, seenwhite;
     int type;
+    int prev_para_type;
     struct stack_item {
        enum {
            stack_nop = 0,             /* do nothing (for error recovery) */
@@ -519,6 +526,11 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
        word **idximplicit;            /* to restore from \u alternatives */
     } *sitem;
     stack parsestk;
+    struct crossparaitem {
+       int type;                      /* currently c_lcont or -1 */
+       int seen_lcont;
+    };
+    stack crossparastk;
     word *indexword, *uword, *iword;
     word *idxwordlist;
     rdstring indexstr;
@@ -530,6 +542,8 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
     macros = newtree234(macrocmp);
     already = FALSE;
 
+    crossparastk = stk_new();
+
     /*
      * Loop on each paragraph.
      */
@@ -542,10 +556,12 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
        /*
         * Get a token.
         */
-       if (!already) {
-           dtor(t), t = get_token(in);
-       }
-       already = FALSE;
+       do {
+           if (!already) {
+               dtor(t), t = get_token(in);
+           }
+           already = FALSE;
+       } while (t.type == tok_eop);
        if (t.type == tok_eof)
            break;
 
@@ -574,18 +590,83 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
                    break;
                else if (t.type != tok_cmd || t.cmd != c_c) {
                    error(err_brokencodepara, &t.pos);
+                   prev_para_type = par.type;
                    addpara(par, ret);
                    while (t.type != tok_eop)   /* error recovery: */
                        dtor(t), t = get_token(in);   /* eat rest of paragraph */
                    goto codeparabroken;   /* ick, but such is life */
                }
            }
+           prev_para_type = par.type;
            addpara(par, ret);
            codeparabroken:
            continue;
        }
 
        /*
+        * Spot the special commands that define a grouping of more
+        * than one paragraph, and also the closing braces that
+        * finish them.
+        */
+       if (t.type == tok_cmd &&
+           t.cmd == c_lcont) {
+           struct crossparaitem *sitem, *stop;
+
+           /*
+            * Expect, and swallow, an open brace.
+            */
+           dtor(t), t = get_token(in);
+           if (t.type != tok_lbrace) {
+               error(err_explbr, &t.pos);
+               continue;
+           }
+
+           /*
+            * \lcont causes a continuation of a list item into
+            * multiple paragraphs (which may in turn contain
+            * nested lists, code paras etc). Hence, the previous
+            * paragraph must be of a list type.
+            */
+           sitem = mknew(struct crossparaitem);
+           stop = (struct crossparaitem *)stk_top(crossparastk);
+           if (prev_para_type == para_Bullet ||
+               prev_para_type == para_NumberedList ||
+               prev_para_type == para_Description) {
+               sitem->type = c_lcont;
+               sitem->seen_lcont = 1;
+               par.type = para_LcontPush;
+               prev_para_type = par.type;
+               addpara(par, ret);
+           } else {
+               /*
+                * Push a null item on the cross-para stack so that
+                * when we see the corresponding closing brace we
+                * don't give a cascade error.
+                */
+               sitem->type = -1;
+               sitem->seen_lcont = (stop ? stop->seen_lcont : 0);
+               error(err_misplacedlcont, &t.pos);
+           }
+           stk_push(crossparastk, sitem);
+           continue;
+       } else if (t.type == tok_rbrace) {
+           struct crossparaitem *sitem = stk_pop(crossparastk);
+           if (!sitem)
+               error(err_unexbrace, &t.pos);
+           else {
+               switch (sitem->type) {
+                 case c_lcont:
+                   par.type = para_LcontPop;
+                   prev_para_type = par.type;
+                   addpara(par, ret);
+                   break;
+               }
+               sfree(sitem);
+           }
+           continue;
+       }
+
+       /*
         * This token begins a paragraph. See if it's one of the
         * special commands that define a paragraph type.
         *
@@ -638,6 +719,8 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
              case c_U: needkw = 32; par.type = para_UnnumberedChapter; break;
                /* For \b and \n the keyword is optional */
              case c_b: needkw = 4; par.type = para_Bullet; break;
+             case c_dt: needkw = 4; par.type = para_DescribedThing; break;
+             case c_dd: needkw = 4; par.type = para_Description; break;
              case c_n: needkw = 4; par.type = para_NumberedList; break;
              case c_cfg: needkw = 8; par.type = para_Config;
                start_cmd = c_cfg; break;
@@ -651,6 +734,17 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
              case c_versionid: needkw = 32; par.type = para_VersionID; break;
            }
 
+           if (par.type == para_Chapter ||
+               par.type == para_Heading ||
+               par.type == para_Subsect ||
+               par.type == para_Appendix ||
+               par.type == para_UnnumberedChapter) {
+               struct crossparaitem *sitem = stk_top(crossparastk);
+               if (sitem && sitem->seen_lcont) {
+                   error(err_sectmarkerinlcont, &t.pos);
+               }
+           }
+
            if (needkw > 0) {
                rdstring rs = { 0, 0, NULL };
                int nkeys = 0;
@@ -732,6 +826,7 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
                    }
                    if (t.type == tok_cmd)
                        already = TRUE;/* inhibit get_token at top of loop */
+                   prev_para_type = par.type;
                    addpara(par, ret);
                    continue;          /* next paragraph */
                }
@@ -837,9 +932,16 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
                break;
              case tok_rbrace:
                sitem = stk_pop(parsestk);
-               if (!sitem)
-                   error(err_unexbrace, &t.pos);
-               else {
+               if (!sitem) {
+                   /*
+                    * This closing brace could have been an
+                    * indication that the cross-paragraph stack
+                    * wants popping. Accordingly, we treat it here
+                    * as an indication that the paragraph is over.
+                    */
+                   already = TRUE;
+                   goto finished_para;
+               } else {
                    if (sitem->type & stack_ualt) {
                        whptr = sitem->whptr;
                        idximplicit = sitem->idximplicit;
@@ -1170,16 +1272,26 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
                dtor(t), t = get_token(in);
            seenwhite = iswhite;
        }
+       finished_para:
        /* Check the stack is empty */
-       if (NULL != (sitem = stk_pop(parsestk))) {
-           do {
+       if (stk_top(parsestk)) {
+           while ((sitem = stk_pop(parsestk)))
                sfree(sitem);
-               sitem = stk_pop(parsestk);
-           } while (sitem);
            error(err_missingrbrace, &t.pos);
        }
        stk_free(parsestk);
+       prev_para_type = par.type;
        addpara(par, ret);
+       if (t.type == tok_eof)
+           already = TRUE;
+    }
+
+    if (stk_top(crossparastk)) {
+       void *p;
+
+       error(err_missingrbrace2, &t.pos);
+       while ((p = stk_pop(crossparastk)))
+           sfree(p);
     }
 
     /*
@@ -1188,6 +1300,8 @@ static void read_file(paragraph ***ret, input *in, indexdata *idx) {
      */
     dtor(t);
     macrocleanup(macros);
+
+    stk_free(crossparastk);
 }
 
 paragraph *read_input(input *in, indexdata *idx) {