summaryrefslogtreecommitdiff
path: root/src/data.c
diff options
context:
space:
mode:
Diffstat (limited to 'src/data.c')
-rw-r--r--src/data.c476
1 files changed, 476 insertions, 0 deletions
diff --git a/src/data.c b/src/data.c
new file mode 100644
index 0000000..dea87b4
--- /dev/null
+++ b/src/data.c
@@ -0,0 +1,476 @@
+/*
+ * Copyright (C) 1989-95 GROUPE BULL
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
+ * GROUPE BULL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ * Except as contained in this notice, the name of GROUPE BULL shall not be
+ * used in advertising or otherwise to promote the sale, use or other dealings
+ * in this Software without prior written authorization from GROUPE BULL.
+ */
+
+/*****************************************************************************\
+* data.c: *
+* *
+* XPM library *
+* IO utilities *
+* *
+* Developed by Arnaud Le Hors *
+\*****************************************************************************/
+/* $XFree86: xc/extras/Xpm/lib/data.c,v 1.4 2002/01/07 19:40:49 dawes Exp $ */
+
+#ifndef CXPMPROG
+#if 0
+/* Official version number */
+static char *RCS_Version = "$XpmVersion: 3.4k $";
+
+/* Internal version number */
+static char *RCS_Id = "Id: xpm.shar,v 3.71 1998/03/19 19:47:14 lehors Exp $";
+#endif
+#include "XpmI.h"
+#endif
+#include <ctype.h>
+
+#ifndef CXPMPROG
+#define Getc(data, file) getc(file)
+#define Ungetc(data, c, file) ungetc(c, file)
+#endif
+
+static int
+ParseComment(xpmData *data)
+{
+ if (data->type == XPMBUFFER) {
+ register char c;
+ register unsigned int n = 0;
+ unsigned int notend;
+ char *s, *s2;
+
+ s = data->Comment;
+ *s = data->Bcmt[0];
+
+ /* skip the string beginning comment */
+ s2 = data->Bcmt;
+ do {
+ c = *data->cptr++;
+ *++s = c;
+ n++;
+ s2++;
+ } while (c == *s2 && *s2 != '\0' && c);
+
+ if (*s2 != '\0') {
+ /* this wasn't the beginning of a comment */
+ data->cptr -= n;
+ return 0;
+ }
+ /* store comment */
+ data->Comment[0] = *s;
+ s = data->Comment;
+ notend = 1;
+ n = 0;
+ while (notend) {
+ s2 = data->Ecmt;
+ while (*s != *s2 && c) {
+ c = *data->cptr++;
+ if (n == XPMMAXCMTLEN - 1) { /* forget it */
+ s = data->Comment;
+ n = 0;
+ }
+ *++s = c;
+ n++;
+ }
+ data->CommentLength = n;
+ do {
+ c = *data->cptr++;
+ if (n == XPMMAXCMTLEN - 1) { /* forget it */
+ s = data->Comment;
+ n = 0;
+ }
+ *++s = c;
+ n++;
+ s2++;
+ } while (c == *s2 && *s2 != '\0' && c);
+ if (*s2 == '\0') {
+ /* this is the end of the comment */
+ notend = 0;
+ data->cptr--;
+ }
+ }
+ return 0;
+ } else {
+ FILE *file = data->stream.file;
+ register int c;
+ register unsigned int n = 0, a;
+ unsigned int notend;
+ char *s, *s2;
+
+ s = data->Comment;
+ *s = data->Bcmt[0];
+
+ /* skip the string beginning comment */
+ s2 = data->Bcmt;
+ do {
+ c = Getc(data, file);
+ *++s = c;
+ n++;
+ s2++;
+ } while (c == *s2 && *s2 != '\0' && c != EOF);
+
+ if (*s2 != '\0') {
+ /* this wasn't the beginning of a comment */
+ /* put characters back in the order that we got them */
+ for (a = n; a > 0; a--, s--)
+ Ungetc(data, *s, file);
+ return 0;
+ }
+ /* store comment */
+ data->Comment[0] = *s;
+ s = data->Comment;
+ notend = 1;
+ n = 0;
+ while (notend) {
+ s2 = data->Ecmt;
+ while (*s != *s2 && c != EOF) {
+ c = Getc(data, file);
+ if (n == XPMMAXCMTLEN - 1) { /* forget it */
+ s = data->Comment;
+ n = 0;
+ }
+ *++s = c;
+ n++;
+ }
+ data->CommentLength = n;
+ do {
+ c = Getc(data, file);
+ if (n == XPMMAXCMTLEN - 1) { /* forget it */
+ s = data->Comment;
+ n = 0;
+ }
+ *++s = c;
+ n++;
+ s2++;
+ } while (c == *s2 && *s2 != '\0' && c != EOF);
+ if (*s2 == '\0') {
+ /* this is the end of the comment */
+ notend = 0;
+ Ungetc(data, *s, file);
+ }
+ }
+ return 0;
+ }
+}
+
+/*
+ * skip to the end of the current string and the beginning of the next one
+ */
+int
+xpmNextString(data)
+ xpmData *data;
+{
+ if (!data->type)
+ data->cptr = (data->stream.data)[++data->line];
+ else if (data->type == XPMBUFFER) {
+ register char c;
+
+ /* get to the end of the current string */
+ if (data->Eos)
+ while ((c = *data->cptr++) && c != data->Eos);
+
+ /*
+ * then get to the beginning of the next string looking for possible
+ * comment
+ */
+ if (data->Bos) {
+ while ((c = *data->cptr++) && c != data->Bos)
+ if (data->Bcmt && c == data->Bcmt[0])
+ ParseComment(data);
+ } else if (data->Bcmt) { /* XPM2 natural */
+ while ((c = *data->cptr++) == data->Bcmt[0])
+ ParseComment(data);
+ data->cptr--;
+ }
+ } else {
+ register int c;
+ FILE *file = data->stream.file;
+
+ /* get to the end of the current string */
+ if (data->Eos)
+ while ((c = Getc(data, file)) != data->Eos && c != EOF);
+
+ /*
+ * then get to the beginning of the next string looking for possible
+ * comment
+ */
+ if (data->Bos) {
+ while ((c = Getc(data, file)) != data->Bos && c != EOF)
+ if (data->Bcmt && c == data->Bcmt[0])
+ ParseComment(data);
+
+ } else if (data->Bcmt) { /* XPM2 natural */
+ while ((c = Getc(data, file)) == data->Bcmt[0])
+ ParseComment(data);
+ Ungetc(data, c, file);
+ }
+ }
+ return 0;
+}
+
+
+/*
+ * skip whitespace and return the following word
+ */
+unsigned int
+xpmNextWord(data, buf, buflen)
+ xpmData *data;
+ char *buf;
+ unsigned int buflen;
+{
+ register unsigned int n = 0;
+ int c;
+
+ if (!data->type || data->type == XPMBUFFER) {
+ while (isspace(c = *data->cptr) && c != data->Eos)
+ data->cptr++;
+ do {
+ c = *data->cptr++;
+ *buf++ = c;
+ n++;
+ } while (!isspace(c) && c != data->Eos && n < buflen);
+ n--;
+ data->cptr--;
+ } else {
+ FILE *file = data->stream.file;
+
+ while ((c = Getc(data, file)) != EOF && isspace(c) && c != data->Eos);
+ while (!isspace(c) && c != data->Eos && c != EOF && n < buflen) {
+ *buf++ = c;
+ n++;
+ c = Getc(data, file);
+ }
+ Ungetc(data, c, file);
+ }
+ return (n);
+}
+
+/*
+ * skip whitespace and compute the following unsigned int,
+ * returns 1 if one is found and 0 if not
+ */
+int
+xpmNextUI(data, ui_return)
+ xpmData *data;
+ unsigned int *ui_return;
+{
+ char buf[BUFSIZ];
+ int l;
+
+ l = xpmNextWord(data, buf, BUFSIZ);
+ return xpmatoui(buf, l, ui_return);
+}
+
+/*
+ * return end of string - WARNING: malloc!
+ */
+int
+xpmGetString(data, sptr, l)
+ xpmData *data;
+ char **sptr;
+ unsigned int *l;
+{
+ unsigned int i, n = 0;
+ int c;
+ char *p = NULL, *q, buf[BUFSIZ];
+
+ if (!data->type || data->type == XPMBUFFER) {
+ if (data->cptr) {
+ char *start = data->cptr;
+ while ((c = *data->cptr) && c != data->Eos)
+ data->cptr++;
+ n = data->cptr - start + 1;
+ p = (char *) XpmMalloc(n);
+ if (!p)
+ return (XpmNoMemory);
+ strncpy(p, start, n);
+ if (data->type) /* XPMBUFFER */
+ p[n - 1] = '\0';
+ }
+ } else {
+ FILE *file = data->stream.file;
+
+ if ((c = Getc(data, file)) == EOF)
+ return (XpmFileInvalid);
+
+ i = 0;
+ q = buf;
+ p = (char *) XpmMalloc(1);
+ while (c != data->Eos && c != EOF) {
+ if (i == BUFSIZ) {
+ /* get to the end of the buffer */
+ /* malloc needed memory */
+ q = (char *) XpmRealloc(p, n + i);
+ if (!q) {
+ XpmFree(p);
+ return (XpmNoMemory);
+ }
+ p = q;
+ q += n;
+ /* and copy what we already have */
+ strncpy(q, buf, i);
+ n += i;
+ i = 0;
+ q = buf;
+ }
+ *q++ = c;
+ i++;
+ c = Getc(data, file);
+ }
+ if (c == EOF) {
+ XpmFree(p);
+ return (XpmFileInvalid);
+ }
+ if (n + i != 0) {
+ /* malloc needed memory */
+ q = (char *) XpmRealloc(p, n + i + 1);
+ if (!q) {
+ XpmFree(p);
+ return (XpmNoMemory);
+ }
+ p = q;
+ q += n;
+ /* and copy the buffer */
+ strncpy(q, buf, i);
+ n += i;
+ p[n++] = '\0';
+ } else {
+ *p = '\0';
+ n = 1;
+ }
+ Ungetc(data, c, file);
+ }
+ *sptr = p;
+ *l = n;
+ return (XpmSuccess);
+}
+
+/*
+ * get the current comment line
+ */
+int
+xpmGetCmt(data, cmt)
+ xpmData *data;
+ char **cmt;
+{
+ if (!data->type)
+ *cmt = NULL;
+ else if (data->CommentLength) {
+ *cmt = (char *) XpmMalloc(data->CommentLength + 1);
+ strncpy(*cmt, data->Comment, data->CommentLength);
+ (*cmt)[data->CommentLength] = '\0';
+ data->CommentLength = 0;
+ } else
+ *cmt = NULL;
+ return 0;
+}
+
+xpmDataType xpmDataTypes[] =
+{
+ {"", "!", "\n", '\0', '\n', "", "", "", ""}, /* Natural type */
+ {"C", "/*", "*/", '"', '"', ",\n", "static char *", "[] = {\n", "};\n"},
+ {"Lisp", ";", "\n", '"', '"', "\n", "(setq ", " '(\n", "))\n"},
+ {NULL, NULL, NULL, 0, 0, NULL, NULL, NULL, NULL}
+};
+
+/*
+ * parse xpm header
+ */
+int
+xpmParseHeader(data)
+ xpmData *data;
+{
+ char buf[BUFSIZ];
+ int l, n = 0;
+
+ if (data->type) {
+ data->Bos = '\0';
+ data->Eos = '\n';
+ data->Bcmt = data->Ecmt = NULL;
+ l = xpmNextWord(data, buf, BUFSIZ);
+ if (l == 7 && !strncmp("#define", buf, 7)) {
+ /* this maybe an XPM 1 file */
+ char *ptr;
+
+ l = xpmNextWord(data, buf, BUFSIZ);
+ if (!l)
+ return (XpmFileInvalid);
+ buf[l] = '\0';
+ ptr = rindex(buf, '_');
+ if (!ptr || strncmp("_format", ptr, l - (ptr - buf)))
+ return XpmFileInvalid;
+ /* this is definitely an XPM 1 file */
+ data->format = 1;
+ n = 1; /* handle XPM1 as mainly XPM2 C */
+ } else {
+
+ /*
+ * skip the first word, get the second one, and see if this is
+ * XPM 2 or 3
+ */
+ l = xpmNextWord(data, buf, BUFSIZ);
+ if ((l == 3 && !strncmp("XPM", buf, 3)) ||
+ (l == 4 && !strncmp("XPM2", buf, 4))) {
+ if (l == 3)
+ n = 1; /* handle XPM as XPM2 C */
+ else {
+ /* get the type key word */
+ l = xpmNextWord(data, buf, BUFSIZ);
+
+ /*
+ * get infos about this type
+ */
+ while (xpmDataTypes[n].type
+ && strncmp(xpmDataTypes[n].type, buf, l))
+ n++;
+ }
+ data->format = 0;
+ } else
+ /* nope this is not an XPM file */
+ return XpmFileInvalid;
+ }
+ if (xpmDataTypes[n].type) {
+ if (n == 0) { /* natural type */
+ data->Bcmt = xpmDataTypes[n].Bcmt;
+ data->Ecmt = xpmDataTypes[n].Ecmt;
+ xpmNextString(data); /* skip the end of the headerline */
+ data->Bos = xpmDataTypes[n].Bos;
+ data->Eos = xpmDataTypes[n].Eos;
+ } else {
+ data->Bcmt = xpmDataTypes[n].Bcmt;
+ data->Ecmt = xpmDataTypes[n].Ecmt;
+ if (!data->format) { /* XPM 2 or 3 */
+ data->Bos = xpmDataTypes[n].Bos;
+ data->Eos = '\0';
+ /* get to the beginning of the first string */
+ xpmNextString(data);
+ data->Eos = xpmDataTypes[n].Eos;
+ } else /* XPM 1 skip end of line */
+ xpmNextString(data);
+ }
+ } else
+ /* we don't know about that type of XPM file... */
+ return XpmFileInvalid;
+ }
+ return XpmSuccess;
+}