Re: test results mc.hlp.it 156+ M

["Andrew V. Samoilov" <sav bcs zp ua>]

Hi, Tribhuvan!
Pavel Tsekov wrote:
> On Tue, 14 Jan 2003, Tribhuvan wrote:
> > I don't have the experience to figure this out as yet.... but if you 
> > have any more ideas for me to try
> > let me know as I'm ready to test further as needed.

Can you test attached patch?  I eliminate ftell()s, it is one of the 
possible bottlenecks.


--
Regards,
Andrew V. Samoilov

--- src/man2hlp.c	Wed Jan 15 16:25:46 2003
+++ src/man2hlp.c	Wed Jan 15 16:34:15 2003
@@ -61,38 +61,9 @@ struct node {
 static struct node nodes;
 static struct node *cnode;	/* Current node */
 
-#define MAX_STREAM_BLOCK 8192
-
-/*
- * Read in blocks of reasonable size and make sure we read everything.
- * Failure to read everything is an error, indicated by returning 0.
- */
-static size_t
-persistent_fread (void *data, size_t len, FILE *stream)
-{
-    size_t count;
-    size_t bytes_done = 0;
-    char *ptr = (char *) data;
-
-    if (len <= 0)
-	return 0;
-
-    while (bytes_done < len) {
-	count = len - bytes_done;
-	if (count > MAX_STREAM_BLOCK)
-	    count = MAX_STREAM_BLOCK;
-
-	count = fread (ptr, 1, count, stream);
-
-	if (count <= 0)
-	    return 0;
+static GString *outfile = NULL;	/* Output file content */
 
-	bytes_done += count;
-	ptr += count;
-    }
-
-    return bytes_done;
-}
+#define MAX_STREAM_BLOCK 8192
 
 /*
  * Write in blocks of reasonable size and make sure we write everything.
@@ -170,7 +141,7 @@ newline (void)
 {
     out_row++;
     col = 0;
-    fprintf (f_out, "\n");
+    g_string_append_c (outfile, '\n');
 }
 
 /* Calculate the length of string */
@@ -236,7 +207,8 @@ print_string (char *buffer)
 		continue;
 	    }
 	    backslash_flag = 0;
-	    fputc (c, f_out);
+
+	    g_string_append_c (outfile, c);
 	}
     } else {
 	/* Split into words */
@@ -251,11 +223,11 @@ print_string (char *buffer)
 		    newline ();
 		/* Words are separated by spaces */
 		if (col > 0) {
-		    fputc (' ', f_out);
+		    g_string_append_c (outfile, ' ');
 		    col++;
 		} else if (indentation) {
 		    while (col++ < indentation)
-			fputc (' ', f_out);
+			g_string_append_c (outfile, ' ');
 		}
 		/* Attempt to handle backslash quoting */
 		while (*(buffer)) {
@@ -265,7 +237,7 @@ print_string (char *buffer)
 			continue;
 		    }
 		    backslash_flag = 0;
-		    fputc (c, f_out);
+		    g_string_append_c (outfile, c);
 		}
 		/* Increase column */
 		col += len;
@@ -336,7 +308,7 @@ handle_node (char *buffer, int is_sh)
 	    if (!is_sh || !node) {
 		/* Start a new section, but omit empty section names */
 		if (*buffer) {
-		    fprintf (f_out, "%c[%s]", CHAR_NODE_END, buffer);
+		    g_string_sprintfa (outfile, "%c[%s]", CHAR_NODE_END, buffer);
 		    col++;
 		    newline ();
 		}
@@ -663,9 +635,6 @@ main (int argc, char **argv)
     FILE *f_tmpl;		/* Template file */
     char buffer[BUFFER_SIZE];	/* Full input line */
     char *node = NULL;
-    char *outfile_buffer;	/* Large buffer to keep the output file */
-    long cont_start;		/* Start of [Contents] */
-    long file_end;		/* Length of the output file */
 
     /* Validity check for arguments */
     if (argc != 4) {
@@ -681,6 +650,7 @@ main (int argc, char **argv)
     /* First stage - process the manual, write to the output file */
     f_man = fopen_check (c_man, "r");
     f_out = fopen_check (c_out, "w");
+    outfile = g_string_sized_new (128*1024);
     c_in = c_man;
 
     /* Repeat for each input line */
@@ -734,7 +704,7 @@ main (int argc, char **argv)
 		    newline ();
 		else
 		    while (++col < indentation)
-			fputc (' ', f_out);
+			g_string_append_c (outfile, ' ');
 	    }
 	}
     }
@@ -782,13 +752,7 @@ main (int argc, char **argv)
 	    } else
 		node = NULL;
 	}
-	fputs (buffer, f_out);
-    }
-
-    cont_start = ftell (f_out);
-    if (cont_start <= 0) {
-	perror (c_out);
-	return 1;
+	g_string_append (outfile, buffer);
     }
 
     if (topics)
@@ -840,53 +804,17 @@ main (int argc, char **argv)
 	cnode = next;
     }
 
-    file_end = ftell (f_out);
-
-    /* Sanity check */
-    if ((file_end <= 0) || (file_end - cont_start <= 0)) {
-	perror (c_out);
-	return 1;
-    }
-
-    fclose_check (f_out);
     fclose_check (f_tmpl);
-    /* Second stage ends here, closing all files, note the end of output */
-
-    /*
-     * Third stage - swap two parts of the output file.
-     * First, open the output file for reading and load it into the memory.
-     */
-    f_out = fopen_check (c_out, "r");
+    /* Second stage ends here */
 
-    outfile_buffer = malloc (file_end);
-    if (!outfile_buffer)
-	return 1;
-
-    if (!persistent_fread (outfile_buffer, file_end, f_out)) {
-	perror (c_out);
-	return 1;
-    }
-
-    fclose_check (f_out);
-    /* Now the output file is in the memory */
-
-    /* Again open output file for writing */
-    f_out = fopen_check (c_out, "w");
-
-    /* Write part after the "Contents" node */
-    if (!persistent_fwrite
-	(outfile_buffer + cont_start, file_end - cont_start, f_out)) {
-	perror (c_out);
-	return 1;
-    }
+    /* Third stage - store output file. */
 
-    /* Write part before the "Contents" node */
-    if (!persistent_fwrite (outfile_buffer, cont_start, f_out)) {
+    if (!persistent_fwrite (outfile->str, outfile->len, f_out)) {
 	perror (c_out);
 	return 1;
     }
 
-    free (outfile_buffer);
+    g_string_free (outfile, TRUE);
     fclose_check (f_out);
     /* Closing everything */