# HG changeset patch
# User ali <ali@juiblex.co.uk>
# Date 1380812979 -3600
# Node ID c45fa3843618a62325745fd3661fd667cff3803d
# Parent  8c2d6a0cf7172aba353cc2bb1bd7fd37c966b663
Fix bug #20: Accept Unicode emdash as valid end-of-paragraph punctuation

diff -r 8c2d6a0cf717 -r c45fa3843618 bookloupe/bookloupe.c
--- a/bookloupe/bookloupe.c	Tue Sep 24 07:18:50 2013 +0100
+++ b/bookloupe/bookloupe.c	Thu Oct 03 16:09:39 2013 +0100
@@ -431,6 +431,40 @@
     return 0;
 }
 
+void count_dashes(const char *line,const char *dash,
+  struct dash_results *results)
+{
+    int i;
+    gchar **tokens;
+    gunichar pc,nc;
+    gboolean spaced=FALSE,unspaced=FALSE,spaced2=FALSE;
+    if (!*line)
+	return;
+    tokens=g_strsplit(line,dash,0);
+    if (tokens[1])
+	results->base++;
+    for(i=1;tokens[i];i++)
+    {
+	pc=g_utf8_get_char(g_utf8_prev_char(tokens[i-1]+strlen(tokens[i-1])));
+	nc=g_utf8_get_char(tokens[i]);
+	if (g_unichar_isspace(pc) || g_unichar_isspace(nc))
+	    spaced=TRUE;
+	if (g_unichar_isspace(pc) && g_unichar_isspace(nc))
+	    spaced2=TRUE;
+	else if (!g_unichar_isspace(pc) && !g_unichar_isspace(nc))
+	    unspaced=TRUE;
+    }
+    if (spaced)
+	results->space++;
+    if (spaced2)
+	/* count of lines with em-dashes with spaces both sides */
+	results->non_PG_space++;
+    if (unspaced)
+	/* count of lines with PG-type em-dashes with no spaces */
+	results->PG_space++;
+    g_strfreev(tokens);
+}
+
 /*
  * first_pass:
  *
@@ -449,6 +483,7 @@
     unsigned int lastlen=0,lastblen=0;
     long spline=0,nspline=0;
     static struct first_pass_results results={0};
+    struct dash_results tmp_dash_results;
     gchar *inword;
     QuoteClass qc;
     lines=g_strsplit(etext,"\n",0);
@@ -512,7 +547,7 @@
 		else
 		    qc=INVALID_QUOTE;
 		if ((qc==CLOSING_QUOTE || qc==NEUTRAL_QUOTE) &&
-		  isalpha(g_utf8_get_char(g_utf8_prev_char(s))))
+		  g_unichar_isalpha(g_utf8_get_char(g_utf8_prev_char(s))))
 		    results.endquote_count++;
 	    }
 	}
@@ -559,18 +594,15 @@
 		results.htmcount+=4; /* bonus marks! */
 	}
 	/* Check for spaced em-dashes */
-	if (lines[j][0] && (s=strstr(g_utf8_next_char(lines[j]),"--")))
-	{
-	    results.emdash++;
-	    if (s[-1]==CHAR_SPACE || s[2]==CHAR_SPACE)
-		results.space_emdash++;
-	    if (s[-1]==CHAR_SPACE && s[2]==CHAR_SPACE)
-		/* count of em-dashes with spaces both sides */
-		results.non_PG_space_emdash++;
-	    if (s[-1]!=CHAR_SPACE && s[2]!=CHAR_SPACE)
-		/* count of PG-type em-dashes with no spaces */
-		results.PG_space_emdash++;
-	}
+	memset(&tmp_dash_results,0,sizeof(tmp_dash_results));
+	count_dashes(lines[j],"--",&tmp_dash_results);
+	count_dashes(lines[j],"—",&tmp_dash_results);
+	if (tmp_dash_results.base)
+	    results.emdash.base++;
+	if (tmp_dash_results.non_PG_space)
+	    results.emdash.non_PG_space++;
+	if (tmp_dash_results.PG_space)
+	    results.emdash.PG_space++;
 	for (s=lines[j];*s;)
 	{
 	    inword=getaword(&s);
@@ -700,13 +732,13 @@
      * and some people insist on them whatever the guidelines say.
      */
     warnings.dash=1;
-    if (results->spacedash+results->non_PG_space_emdash>
-      results->PG_space_emdash)
+    if (results->spacedash+results->emdash.non_PG_space>
+      results->emdash.PG_space)
     {
 	warnings.dash=0;
 	g_print("   --> There are %ld spaced dashes and em-dashes. "
 	  "Not reporting them.\n",
-	  results->spacedash+results->non_PG_space_emdash);
+	  results->spacedash+results->emdash.non_PG_space);
     }
     /* If more than a quarter of characters are hi-bit, bug out. */
     warnings.bin=1;
@@ -1120,20 +1152,53 @@
 }
 
 /*
+ * str_emdash:
+ *
+ * Find the first em-dash, return a pointer to it and set <next> to the
+ * character following the dash.
+ */
+char *str_emdash(const char *s,const char **next)
+{
+    const char *s1,*s2;
+    s1=strstr(s,"--");
+    s2=strstr(s,"—");
+    if (!s1)
+    {
+	if (s2)
+	    *next=g_utf8_next_char(s2);
+	return (char *)s2;
+    }
+    else if (!s2)
+    {
+	*next=g_utf8_next_char(g_utf8_next_char(s1));
+	return (char *)s1;
+    }
+    else if (s1<s2)
+    {
+	*next=g_utf8_next_char(g_utf8_next_char(s1));
+	return (char *)s1;
+    }
+    else
+    {
+	*next=g_utf8_next_char(s2);
+	return (char *)s2;
+    }
+}
+
+/*
  * check_for_spaced_emdash:
  *
  * Check for spaced em-dashes.
  *
- * We must check _all_ occurrences of "--" on the line
- * hence the loop - even if the first double-dash is OK
+ * We must check _all_ occurrences of em-dashes on the line
+ * hence the loop - even if the first dash is OK
  * there may be another that's wrong later on.
  */
 void check_for_spaced_emdash(const char *aline)
 {
     const char *s,*t,*next;
-    for (s=aline;t=strstr(s,"--");s=next)
+    for (s=aline;t=str_emdash(s,&next);s=next)
     {
-	next=g_utf8_next_char(g_utf8_next_char(t));
 	if (t>aline && g_utf8_get_char(g_utf8_prev_char(t))==CHAR_SPACE ||
 	  g_utf8_get_char(next)==CHAR_SPACE)
 	{
@@ -2322,7 +2387,7 @@
 	qc=CHAR_IS_DQUOTE(c)?QUOTE_CLASS(c):INVALID_QUOTE;
 	nc=g_utf8_get_char(g_utf8_next_char(s));
 	/* for each character in the line except 1st */
-	if ((qc==CLOSING_QUOTE || qc==NEUTRAL_QUOTE) && isalpha(pc))
+	if ((qc==CLOSING_QUOTE || qc==NEUTRAL_QUOTE) && g_unichar_isalpha(pc))
 	{
 	    if (pswit[ECHO_SWITCH])
 		g_print("\n%s\n",aline);
@@ -2466,7 +2531,7 @@
 		    cnt_punct++;
 		break;
 	    }
-	    if (g_utf8_strchr("-.:!([{?}])",-1,g_utf8_get_char(s)))
+	    if (g_utf8_strchr("—-.:!([{?}])",-1,g_utf8_get_char(s)))
 		break;
 	}
     }
diff -r 8c2d6a0cf717 -r c45fa3843618 bookloupe/bookloupe.h
--- a/bookloupe/bookloupe.h	Tue Sep 24 07:18:50 2013 +0100
+++ b/bookloupe/bookloupe.h	Thu Oct 03 16:09:39 2013 +0100
@@ -58,11 +58,16 @@
     SWITNO
 };
 
+struct dash_results {
+    long base,space,non_PG_space,PG_space;
+};
+
 struct first_pass_results {
     long firstline,astline;
     long footerline,totlen,binlen,alphalen,endquote_count,shortline,dotcomma;
     long fslashline,hyphens,longline,verylongline,htmcount,standalone_digit;
-    long spacedash,emdash,space_emdash,non_PG_space_emdash,PG_space_emdash;
+    long spacedash;
+    struct dash_results emdash;
     int Dutchcount,Frenchcount;
 };
 
diff -r 8c2d6a0cf717 -r c45fa3843618 test/bookloupe/Makefile.am
--- a/test/bookloupe/Makefile.am	Tue Sep 24 07:18:50 2013 +0100
+++ b/test/bookloupe/Makefile.am	Thu Oct 03 16:09:39 2013 +0100
@@ -1,5 +1,6 @@
 TESTS_ENVIRONMENT=BOOKLOUPE=../../bookloupe/bookloupe ../harness/loupe-test
 TESTS=non-ascii.tst long-line.tst curved-single-quotes.tst curved-quotes.tst \
-	runfox-quotes.tst curved-genitives.tst multi-line-illustration.tst
+	runfox-quotes.tst curved-genitives.tst multi-line-illustration.tst \
+	emdash.tst
 
 dist_pkgdata_DATA=$(TESTS)
diff -r 8c2d6a0cf717 -r c45fa3843618 test/bookloupe/emdash.tst
--- /dev/null	Thu Jan 01 00:00:00 1970 +0000
+++ b/test/bookloupe/emdash.tst	Thu Oct 03 16:09:39 2013 +0100
@@ -0,0 +1,14 @@
+**************** INPUT ****************
+“Those are good words,” Spotted Deer declared, admiringly. “You will
+soon find Gokhos, the great white Medicine Owl, and then we will be able
+to do some great things. I am not thinking about the Shawnees——”
+
+“Stop!” Running Fox interrupted, excitedly. “I see smoke rising behind
+that ridge — I believe we have found the Mohawk camp.”
+**************** WARNINGS ****************
+<expected>
+  <error>
+    <at line="6" column="12"/>
+    <text>Spaced em-dash?</text>
+  </error>
+</expected>