@@ -1,5 +1,5 @@
11#include "hexml.h"
2−#include <malloc.h>
2+#include <stdlib.h>
33#include <string.h>
44#include <assert.h>
55#include <ctype.h>
@@ -10,18 +10,18 @@ typedef int bool;
1010/////////////////////////////////////////////////////////////////////
1111// TYPES
1212
13−int static inline end(str s) { return s.start + s.length; }
13+static inline int end(str s) { return s.start + s.length; }
1414
15−str static inline start_length(int32_t start, int32_t length)
15+static inline str start_length(int32_t start, int32_t length)
1616{
17− if (start < 0 || length < 0) assert(0);
17+ assert(start >= 0 && length >= 0);
1818 str res;
1919 res.start = start;
2020 res.length = length;
2121 return res;
2222}
2323
24−str static inline start_end(int32_t start, int32_t end)
24+static inline str start_end(int32_t start, int32_t end)
2525{
2626 return start_length(start, end - start);
2727}
@@ -40,7 +40,7 @@ typedef struct
4040 // have a cursor at the end, which is stack scoped, so children write out, then I do
4141 // when you commit, you copy over from end to front
4242
43− // nodes
43+ // nodes
4444 int size;
4545 int used_front; // front entries, stored for good
4646 int used_back; // back entries, stack based, copied into front
@@ -51,11 +51,11 @@ typedef struct
5151
5252struct document
5353{
54− char* body; // pointer to initial argument, not owned by us
54+ const char* body; // pointer to initial argument, not owned by us
5555
5656 // things only used while parsing
57− char* cursor; // pointer to where we are in body
58− char* end; // pointer to one past the last char
57+ const char* cursor; // pointer to where we are in body
58+ const char* end; // pointer to one past the last char
5959 // if cursor is > end we have gone past the end
6060
6161 char* error_message;
@@ -63,8 +63,8 @@ struct document
6363 attr_buffer attrs;
6464};
6565
66−int static inline doc_length(document* d) { return d->end - d->body; }
67−int static inline doc_position(document* d) { return d->cursor - d->body; }
66+static inline int doc_length(const document* d) { return (int) (d->end - d->body); }
67+static inline int doc_position(const document* d) { return (int) (d->cursor - d->body); }
6868
6969
7070/////////////////////////////////////////////////////////////////////
@@ -72,49 +72,45 @@ int static inline doc_position(document* d) { return d->cursor - d->body; }
7272
7373typedef struct
7474{
75− document* d;
75+ const document* d;
7676 char* buffer;
7777 int length;
7878 int cursor;
7979} render;
8080
81−void static inline render_char(render* r, char c)
81+static inline void render_char(render* r, char c)
8282{
8383 if (r->cursor < r->length)
8484 r->buffer[r->cursor] = c;
8585 r->cursor++;
8686}
8787
88−void static inline bound(char* msg, int index, int mn, int mx)
88+static inline void bound(int idx, int mn, int mx)
8989{
90− if (index < mn || index > mx)
91− {
92− //printf("Bounds checking failed %s, got %i, which should be in %i:%i\n", msg, index, mn, mx);
93− assert(0);
94− }
90+ assert(idx >= mn && idx <= mx);
9591}
9692
97−void static inline bound_str(char* msg, str s, int mn, int mx)
93+static inline void bound_str(str s, int mn, int mx)
9894{
99− if (s.length < 0) assert(0);
100− bound(msg, s.start, mn, mx);
101− bound(msg, end(s), mn, mx);
95+ assert(s.length >= 0);
96+ bound(s.start, mn, mx);
97+ bound(end(s), mn, mx);
10298}
10399
104−void render_str(render* r, str s)
100+static void render_str(render* r, str s)
105101{
106− bound_str("render_str", s, 0, doc_length(r->d));
102+ bound_str(s, 0, doc_length(r->d));
107103 for (int i = 0; i < s.length; i++)
108104 render_char(r, r->d->body[s.start + i]);
109105}
110106
111−void render_tag(render* r, node* n);
107+static void render_tag(render* r, const node* n);
112108
113−void render_content(render* r, node* n)
109+static void render_content(render* r, const node* n)
114110{
115− bound_str("render_conent inner", n->inner, 0, doc_length(r->d));
116− bound_str("render_conent nodes", n->nodes, 0, r->d->nodes.used_front);
117− bound_str("render_conent attrs", n->attrs, 0, r->d->attrs.used);
111+ bound_str(n->inner, 0, doc_length(r->d));
112+ bound_str(n->nodes, 0, r->d->nodes.used_front);
113+ bound_str(n->attrs, 0, r->d->attrs.used);
118114
119115 int done = n->inner.start;
120116 for (int i = 0; i < n->nodes.length; i++)
@@ -127,7 +123,7 @@ void render_content(render* r, node* n)
127123 render_str(r, start_end(done, end(n->inner)));
128124}
129125
130−void render_tag(render* r, node* n)
126+static void render_tag(render* r, const node* n)
131127{
132128 render_char(r, '<');
133129 render_str(r, n->name);
@@ -149,7 +145,7 @@ void render_tag(render* r, node* n)
149145 render_char(r, '>');
150146}
151147
152−int node_render(document* d, node* n, char* buffer, int length)
148+int hexml_node_render(const document* d, const node* n, char* buffer, int length)
153149{
154150 render r;
155151 r.d = d;
@@ -168,10 +164,10 @@ int node_render(document* d, node* n, char* buffer, int length)
168164/////////////////////////////////////////////////////////////////////
169165// NOT THE PARSER
170166
171−char* document_error(document* d){return d->error_message;}
172−node* document_node(document* d){return &d->nodes.nodes[0];}
167+char* hexml_document_error(const document* d){return d->error_message;}
168+node* hexml_document_node(const document* d){return &d->nodes.nodes[0];}
173169
174−void document_free(document* d)
170+void hexml_document_free(document* d)
175171{
176172 free(d->error_message);
177173 free(d->nodes.alloc);
@@ -179,25 +175,24 @@ void document_free(document* d)
179175 free(d);
180176}
181177
182−node* node_children(document* d, node* n, int* res)
178+node* hexml_node_children(const document* d, const node* n, int* res)
183179{
184180 *res = n->nodes.length;
185181 return &d->nodes.nodes[n->nodes.start];
186182}
187183
188−attr* node_attributes(document* d, node* n, int* res)
184+attr* hexml_node_attributes(const document* d, const node* n, int* res)
189185{
190186 *res = n->attrs.length;
191187 return &d->attrs.attrs[n->attrs.start];
192188}
193189
194190
195−attr* node_attributeBy(document* d, node* n, char* s, int slen)
191+attr* hexml_node_attribute(const document* d, const node* n, const char* s, int slen)
196192{
197− if (slen == -1) slen = strlen(s);
198− int i = n->attrs.start;
199− int limit = end(n->attrs);
200− for (i = n->attrs.start; i < limit; i++)
193+ if (slen == -1) slen = (int) strlen(s);
194+ const int limit = end(n->attrs);
195+ for (int i = n->attrs.start; i < limit; i++)
201196 {
202197 attr* r = &d->attrs.attrs[i];
203198 if (r->name.length == slen && memcmp(s, &d->body[r->name.start], slen) == 0)
@@ -207,11 +202,11 @@ attr* node_attributeBy(document* d, node* n, char* s, int slen)
207202}
208203
209204// Search for given strings within a node
210−node* node_childBy(document* d, node* parent, node* prev, char* s, int slen)
205+node* hexml_node_child(const document* d, const node* parent, const node* prev, const char* s, int slen)
211206{
212− if (slen == -1) slen = strlen(s);
213− int i = prev == NULL ? parent->nodes.start : prev + 1 - d->nodes.nodes;
214− int limit = end(parent->nodes);
207+ if (slen == -1) slen = (int) strlen(s);
208+ int i = prev == NULL ? parent->nodes.start : (int) (prev + 1 - d->nodes.nodes);
209+ const int limit = end(parent->nodes);
215210 for (; i < limit; i++)
216211 {
217212 node* r = &d->nodes.nodes[i];
@@ -225,49 +220,49 @@ node* node_childBy(document* d, node* parent, node* prev, char* s, int slen)
225220/////////////////////////////////////////////////////////////////////
226221// PARSE TABLE
227222
228−const char tag_name1 = 0x1;
229−const char tag_name = 0x2;
230−const char tag_space = 0x4;
223+static const char tag_name1 = 0x1;
224+static const char tag_name = 0x2;
225+static const char tag_space = 0x4;
231226
232−char parse_table[256];
227+static char parse_table[256];
233228
234−void init_parse_table()
229+static void init_parse_table()
235230{
236231 static bool done = 0;
237232 if (done) return;
238233 for (int i = 0; i < 256; i++)
239234 {
240235 bool name1 = i == ':' || i == '_' || (i >= 'a' && i <= 'z') || (i >= 'A' && i <= 'Z');
241− bool name = name1 || i == '-' || (i >= '0' && i <= '9');
236+ bool name = name1 || i == '-' || i == '.' || (i >= '0' && i <= '9');
242237 bool space = i == ' ' || i == '\t' || i == '\r' || i == '\n';
243238 parse_table[i] = (name1 ? tag_name1 : 0) | (name ? tag_name : 0) | (space ? tag_space : 0);
244239 }
245240 done = 1;
246241}
247242
248−bool static inline is(char c, char tag) { return parse_table[c] & tag; }
249−bool static inline is_name1(char c) { return is(c, tag_name1); }
250−bool static inline is_name(char c) { return is(c, tag_name); }
251−bool static inline is_space(char c) { return is(c, tag_space); }
243+static inline bool is(char c, char tag) { return parse_table[(unsigned char) c] & tag; }
244+static inline bool is_name1(char c) { return is(c, tag_name1); }
245+static inline bool is_name(char c) { return is(c, tag_name); }
246+static inline bool is_space(char c) { return is(c, tag_space); }
252247
253248
254249/////////////////////////////////////////////////////////////////////
255250// PARSER COMBINATORS
256251
257−char static inline peekAt(document* d, int i) { return d->cursor[i]; }
258−void static inline skip(document* d, int i) { d->cursor += i; }
259−char static inline peek(document* d) { return peekAt(d, 0); }
260−char static inline get(document* d) { char c = peek(d); skip(d, 1); return c; }
252+static inline char peek_at(const document* d, int i) { return d->cursor[i]; }
253+static inline char peek(const document* d) { return peek_at(d, 0); }
254+static inline void skip(document* d, int i) { d->cursor += i; }
255+static inline char get(document* d) { char c = peek(d); skip(d, 1); return c; }
261256
262257// Remove whitespace characters from the cursor while they are still whitespace
263−void static inline trim(document* d)
258+static inline void trim(document* d)
264259{
265260 while (is_space(peek(d)))
266261 skip(d, 1);
267262}
268263
269264// Find this character form the cursor onwards, if true adjust the cursor to that char, otherwise leave it at the end
270−bool find(document* d, char c)
265+static bool find(document* d, char c)
271266{
272267 char* x = memchr(d->cursor, c, d->end - d->cursor);
273268 if (x == NULL)
@@ -286,35 +281,46 @@ bool find(document* d, char c)
286281/////////////////////////////////////////////////////////////////////
287282// PARSING CODE
288283
289−void static inline node_alloc(node_buffer* b, int ask)
284+static inline void node_alloc(node_buffer* b, int ask)
290285{
291286 int space = b->size - b->used_back - b->used_front;
292287 if (space >= ask) return;
293− assert(0);
288+ int size2 = (b->size + 1000 + ask) * 2;
289+ node* buf2 = malloc(size2 * sizeof(node));
290+ assert(buf2);
291+ memcpy(buf2, b->nodes, b->used_front * sizeof(node));
292+ memcpy(&buf2[size2 - b->used_back], &b->nodes[b->size - b->used_back], b->used_back * sizeof(node));
293+ free(b->alloc);
294+ b->size = size2;
295+ b->nodes = buf2;
296+ b->alloc = buf2;
294297}
295298
296−void static inline attr_alloc(attr_buffer* b, int ask)
299+static inline void attr_alloc(attr_buffer* b, int ask)
297300{
298301 int space = b->size - b->used;
299302 if (space >= ask) return;
300303 int size2 = (b->size + 1000 + ask) * 2;
301304 attr* buf2 = malloc(size2 * sizeof(attr));
305+ assert(buf2);
302306 memcpy(buf2, b->attrs, b->used * sizeof(attr));
303307 free(b->alloc);
308+ b->size = size2;
304309 b->attrs = buf2;
305310 b->alloc = buf2;
306311}
307312
308−void set_error(document* d, char* msg)
313+static void set_error(document* d, const char* msg)
309314{
310315 if (d->error_message != NULL) return; // keep the first error message
311316 d->error_message = malloc(strlen(msg)+1);
317+ assert(d->error_message);
312318 strcpy(d->error_message, msg);
313319}
314320
315321// you now expect a name, perhaps preceeded by whitespace
316322// the name may be empty
317−str static inline parse_name(document* d)
323+static inline str parse_name(document* d)
318324{
319325 int start = doc_position(d);
320326 if (!is_name1(peek(d)))
@@ -325,14 +331,15 @@ str static inline parse_name(document* d)
325331 return start_end(start, doc_position(d));
326332}
327333
328−str static inline parse_attrval(document* d)
334+static inline str parse_attrval(document* d)
329335{
330336 trim(d);
331− if (get(d) != '=')
337+ if (peek(d) != '=')
332338 {
333339 set_error(d, "Expected = in attribute, but missing");
334340 return start_length(0, 0);
335341 }
342+ skip(d, 1);
336343 trim(d);
337344 char c = peek(d);
338345 if (c == '\"' || c == '\'')
@@ -358,11 +365,10 @@ str static inline parse_attrval(document* d)
358365
359366// seen a tag name, now looking for attributes terminated by >
360367// puts the attributes it finds in the attribute buffer
361−str static inline parse_attributes(document* d)
368+static inline str parse_attributes(document* d)
362369{
363− str res;
364− res.start = d->attrs.used;
365− for (int i = 0; ; i++)
370+ int start = d->attrs.used;
371+ while (d->error_message == NULL)
366372 {
367373 trim(d);
368374 str name = parse_name(d);
@@ -372,34 +378,39 @@ str static inline parse_attributes(document* d)
372378 d->attrs.attrs[d->attrs.used].value = parse_attrval(d);
373379 d->attrs.used++;
374380 }
375− res.length = d->attrs.used - res.start;
376− return res;
381+ return start_end(start, d->attrs.used);
377382}
378383
379−str parse_content(document* d);
384+static str parse_content(document* d);
380385
381386
382387// Add a new entry into tag, am at a '<'
383−void static inline parse_tag(document* d)
388+static inline void parse_tag(document* d)
384389{
385390 node_alloc(&d->nodes, 1);
386391 d->nodes.used_back++;
387− int me = d->nodes.size - d->nodes.used_back;
392+ node* me = &d->nodes.nodes[d->nodes.size - d->nodes.used_back];
388393
389− d->nodes.nodes[me].outer.start = doc_position(d);
394+ me->outer.start = doc_position(d);
390395 char c = get(d);
391396 assert(c == '<');
392397 if (peek(d) == '?') skip(d, 1);
393− d->nodes.nodes[me].name = parse_name(d);
394− d->nodes.nodes[me].attrs = parse_attributes(d);
398+ me->name = parse_name(d);
399+ if (me->name.length == 0)
400+ {
401+ set_error(d, "Missing tag name");
402+ return;
403+ }
404+ me->attrs = parse_attributes(d);
405+ if (d->error_message != NULL) return;
395406
396− c = get(d);
397− if ((c == '/' || c == '?') && peek(d) == '>')
407+ c = peek(d);
408+ if ((c == '/' || c == '?') && peek_at(d, 1) == '>')
398409 {
399− skip(d, 1);
400− d->nodes.nodes[me].nodes = start_length(0, 0);
401− d->nodes.nodes[me].outer.length = start_end(d->nodes.nodes[me].outer.start, doc_position(d)).length;
402− d->nodes.nodes[me].inner = start_length(doc_position(d), 0);
410+ skip(d, 2);
411+ me->nodes = start_length(0, 0);
412+ me->outer.length = start_end(me->outer.start, doc_position(d)).length;
413+ me->inner = start_length(doc_position(d), 0);
403414 return;
404415 }
405416 else if (c != '>')
@@ -407,22 +418,28 @@ void static inline parse_tag(document* d)
407418 set_error(d, "Gunk at the end of the tag");
408419 return;
409420 }
410− d->nodes.nodes[me].inner.start = doc_position(d);
411− d->nodes.nodes[me].nodes = parse_content(d);
412− d->nodes.nodes[me].inner.length = start_end(d->nodes.nodes[me].inner.start, doc_position(d)).length;
421+ else
422+ skip(d, 1);
423+ me->inner.start = doc_position(d);
424+ str content = parse_content(d);
425+
426+ // parse_content may have allocated more nodes, so recompute me
427+ me = &d->nodes.nodes[d->nodes.size - d->nodes.used_back];
428+ me->nodes = content;
429+ me->inner.length = start_end(me->inner.start, doc_position(d)).length;
413430
414431 if (d->error_message != NULL) return;
415− if (peek(d) == '<' && peekAt(d, 1) == '/')
432+ if (peek(d) == '<' && peek_at(d, 1) == '/')
416433 {
417434 skip(d, 2);
418− if (d->end - d->cursor >= d->nodes.nodes[me].name.length &&
419− memcmp(d->cursor, &d->body[d->nodes.nodes[me].name.start], d->nodes.nodes[me].name.length) == 0)
435+ if (d->end - d->cursor >= me->name.length &&
436+ memcmp(d->cursor, &d->body[me->name.start], me->name.length) == 0)
420437 {
421− skip(d, d->nodes.nodes[me].name.length);
438+ skip(d, me->name.length);
422439 trim(d);
423440 if (get(d) == '>')
424441 {
425− d->nodes.nodes[me].outer.length = start_end(d->nodes.nodes[me].outer.start, doc_position(d)).length;
442+ me->outer.length = start_end(me->outer.start, doc_position(d)).length;
426443 return;
427444 }
428445 }
@@ -434,7 +451,7 @@ void static inline parse_tag(document* d)
434451
435452// Parser until </, return the index of your node children
436453// Not inline as it is recursive
437−str parse_content(document* d)
454+static str parse_content(document* d)
438455{
439456 int before = d->nodes.used_back;
440457 while (d->error_message == NULL)
@@ -444,22 +461,22 @@ str parse_content(document* d)
444461 {
445462 break;
446463 }
447− else if (peekAt(d, 1) == '/')
464+ else if (peek_at(d, 1) == '/')
448465 {
449466 // have found a </
450467 break;
451468 }
452− else if (peekAt(d, 1) == '!' && peekAt(d, 2) == '-' && peekAt(d, 3) == '-')
469+ else if (peek_at(d, 1) == '!' && peek_at(d, 2) == '-' && peek_at(d, 3) == '-')
453470 {
454471 skip(d, 3);
455472 // you can't reuse the two '-' characters for the closing as well
456− if (peekAt(d, 0) == '\0' || peekAt(d, 1) == '\0')
473+ if (peek_at(d, 0) == '\0' || peek_at(d, 1) == '\0')
457474 {
458475 set_error(d, "Didn't get a closing comment");
459476 return start_end(0, 0);
460477 }
461478 skip(d, 2);
462− while (1)
479+ for (;;)
463480 {
464481 if (!find(d, '>'))
465482 {
@@ -467,7 +484,7 @@ str parse_content(document* d)
467484 return start_end(0, 0);
468485 }
469486 skip(d, 1);
470− if (peekAt(d, -3) == '-' && peekAt(d, -2))
487+ if (peek_at(d, -3) == '-' && peek_at(d, -2) == '-')
471488 break;
472489 }
473490 }
@@ -486,20 +503,23 @@ str parse_content(document* d)
486503 return res;
487504}
488505
506+// Based on looking at ~50Kb XML documents, they seem to have ~700 attributes
507+// and ~300 nodes, so size appropriately to cope with that.
489508typedef struct
490509{
491510 document document;
492511 attr attrs[1000];
493− node nodes[1000];
512+ node nodes[500];
494513} buffer;
495514
496−document* document_parse(char* s, int slen)
515+document* hexml_document_parse(const char* s, int slen)
497516{
498− if (slen == -1) slen = strlen(s);
517+ if (slen == -1) slen = (int) strlen(s);
499518 assert(s[slen] == 0);
500519 init_parse_table();
501520
502521 buffer* buf = malloc(sizeof(buffer));
522+ assert(buf);
503523 document* d = &buf->document;
504524 d->body = s;
505525 d->cursor = s;
@@ -509,7 +529,7 @@ document* document_parse(char* s, int slen)
509529 d->attrs.used = 0;
510530 d->attrs.attrs = buf->attrs;
511531 d->attrs.alloc = NULL;
512− d->nodes.size = 1000;
532+ d->nodes.size = 500;
513533 d->nodes.used_back = 0;
514534 d->nodes.used_front = 1;
515535 d->nodes.nodes = buf->nodes;
@@ -519,11 +539,13 @@ document* document_parse(char* s, int slen)
519539 d->nodes.nodes[0].outer = start_length(0, slen);
520540 d->nodes.nodes[0].inner = start_length(0, slen);
521541 d->nodes.nodes[0].attrs = start_length(0, 0);
522− d->nodes.nodes[0].nodes = parse_content(d);
523542
524− if (d->cursor < d->end && d->error_message == NULL)
525− {
543+ // Introduce an intermediate result, otherwise behaviour is undefined
544+ // because there is no guaranteed ordering between LHS and RHS evaluation
545+ str content = parse_content(d);
546+ d->nodes.nodes[0].nodes = content;
547+
548+ if (d->cursor < d->end)
526549 set_error(d, "Trailing junk at the end of the document");
527− }
528550 return d;
529551}