@@ -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);
@@ -149,14 +145,18 @@ void render_tag(render* r, node* n)
149145 render_char(r, '>');
150146}
151147
152−int document_render(document* d, 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;
156152 r.buffer = buffer;
157153 r.length = length;
158154 r.cursor = 0;
159− render_content(&r, document_node(d));
155+ // The root node (and only the root node) has an empty length, so just render its innards
156+ if (n->name.length == 0)
157+ render_content(&r, n);
158+ else
159+ render_tag(&r, n);
160160 return r.cursor;
161161}
162162
@@ -164,10 +164,10 @@ int document_render(document* d, char* buffer, int length)
164164/////////////////////////////////////////////////////////////////////
165165// NOT THE PARSER
166166
167−char* document_error(document* d){return d->error_message;}
168−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];}
169169
170−void document_free(document* d)
170+void hexml_document_free(document* d)
171171{
172172 free(d->error_message);
173173 free(d->nodes.alloc);
@@ -175,93 +175,94 @@ void document_free(document* d)
175175 free(d);
176176}
177177
178−node* node_children(document* d, node* n, int* res)
178+node* hexml_node_children(const document* d, const node* n, int* res)
179179{
180180 *res = n->nodes.length;
181181 return &d->nodes.nodes[n->nodes.start];
182182}
183183
184−attr* node_attributes(document* d, node* n, int* res)
184+attr* hexml_node_attributes(const document* d, const node* n, int* res)
185185{
186186 *res = n->attrs.length;
187187 return &d->attrs.attrs[n->attrs.start];
188188}
189189
190190
191−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)
192192{
193− assert(0);
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++)
196+ {
197+ attr* r = &d->attrs.attrs[i];
198+ if (r->name.length == slen && memcmp(s, &d->body[r->name.start], slen) == 0)
199+ return r;
200+ }
194201 return NULL;
195202}
196203
197204// Search for given strings within a node
198−node* node_firstChildBy(document* d, node* n, char* s, int slen)
205+node* hexml_node_child(const document* d, const node* parent, const node* prev, const char* s, int slen)
199206{
200− if (slen == -1) slen = strlen(s);
201− int i = n->nodes.start;
202− int limit = i + n->nodes.length;
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);
203210 for (; i < limit; i++)
204211 {
205− if (d->nodes.nodes[i].name.length == slen &&
206− memcmp(s, &d->body[d->nodes.nodes[i].name.start], slen) == 0)
207− return &d->nodes.nodes[i];
212+ node* r = &d->nodes.nodes[i];
213+ if (r->name.length == slen && memcmp(s, &d->body[r->name.start], slen) == 0)
214+ return r;
208215 }
209216 return NULL;
210217}
211218
212−node* node_nextChildBy(document* d, node* parent, node* prev, char* s, int slen)
213−{
214− assert(0);
215− return NULL;
216−}
217−
218219
219220/////////////////////////////////////////////////////////////////////
220221// PARSE TABLE
221222
222−const char tag_name1 = 0x1;
223−const char tag_name = 0x2;
224−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;
225226
226−char parse_table[256];
227+static char parse_table[256];
227228
228−void init_parse_table()
229+static void init_parse_table()
229230{
230231 static bool done = 0;
231232 if (done) return;
232233 for (int i = 0; i < 256; i++)
233234 {
234235 bool name1 = i == ':' || i == '_' || (i >= 'a' && i <= 'z') || (i >= 'A' && i <= 'Z');
235− bool name = name1 || i == '-' || (i >= '0' && i <= '9');
236+ bool name = name1 || i == '-' || i == '.' || (i >= '0' && i <= '9');
236237 bool space = i == ' ' || i == '\t' || i == '\r' || i == '\n';
237238 parse_table[i] = (name1 ? tag_name1 : 0) | (name ? tag_name : 0) | (space ? tag_space : 0);
238239 }
239240 done = 1;
240241}
241242
242−bool static inline is(char c, char tag) { return parse_table[c] & tag; }
243−bool static inline is_name1(char c) { return is(c, tag_name1); }
244−bool static inline is_name(char c) { return is(c, tag_name); }
245−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); }
246247
247248
248249/////////////////////////////////////////////////////////////////////
249250// PARSER COMBINATORS
250251
251−char static inline peekAt(document* d, int i) { return d->cursor[i]; }
252−void static inline skip(document* d, int i) { d->cursor += i; }
253−char static inline peek(document* d) { return peekAt(d, 0); }
254−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; }
255256
256257// Remove whitespace characters from the cursor while they are still whitespace
257−void static inline trim(document* d)
258+static inline void trim(document* d)
258259{
259260 while (is_space(peek(d)))
260261 skip(d, 1);
261262}
262263
263264// Find this character form the cursor onwards, if true adjust the cursor to that char, otherwise leave it at the end
264−bool find(document* d, char c)
265+static bool find(document* d, char c)
265266{
266267 char* x = memchr(d->cursor, c, d->end - d->cursor);
267268 if (x == NULL)
@@ -280,37 +281,47 @@ bool find(document* d, char c)
280281/////////////////////////////////////////////////////////////////////
281282// PARSING CODE
282283
283−void static inline node_alloc(node_buffer* b, int ask)
284+static inline void node_alloc(node_buffer* b, int ask)
284285{
285286 int space = b->size - b->used_back - b->used_front;
286287 if (space >= ask) return;
287− 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;
288297}
289298
290−void static inline attr_alloc(attr_buffer* b, int ask)
299+static inline void attr_alloc(attr_buffer* b, int ask)
291300{
292301 int space = b->size - b->used;
293302 if (space >= ask) return;
294303 int size2 = (b->size + 1000 + ask) * 2;
295304 attr* buf2 = malloc(size2 * sizeof(attr));
305+ assert(buf2);
296306 memcpy(buf2, b->attrs, b->used * sizeof(attr));
297307 free(b->alloc);
308+ b->size = size2;
298309 b->attrs = buf2;
299310 b->alloc = buf2;
300311}
301312
302−void set_error(document* d, char* msg)
313+static void set_error(document* d, const char* msg)
303314{
304315 if (d->error_message != NULL) return; // keep the first error message
305316 d->error_message = malloc(strlen(msg)+1);
317+ assert(d->error_message);
306318 strcpy(d->error_message, msg);
307319}
308320
309321// you now expect a name, perhaps preceeded by whitespace
310322// the name may be empty
311−str static inline parse_name(document* d)
323+static inline str parse_name(document* d)
312324{
313− trim(d);
314325 int start = doc_position(d);
315326 if (!is_name1(peek(d)))
316327 return start_length(start, 0);
@@ -320,10 +331,14 @@ str static inline parse_name(document* d)
320331 return start_end(start, doc_position(d));
321332}
322333
323−str static inline parse_attrval(document* d)
334+static inline str parse_attrval(document* d)
324335{
325336 trim(d);
326− if (peek(d) != '=') return start_length(0,0);
337+ if (peek(d) != '=')
338+ {
339+ set_error(d, "Expected = in attribute, but missing");
340+ return start_length(0, 0);
341+ }
327342 skip(d, 1);
328343 trim(d);
329344 char c = peek(d);
@@ -340,19 +355,22 @@ str static inline parse_attrval(document* d)
340355 return start_end(start, doc_position(d) - 1);
341356 }
342357 else
343− return parse_name(d);
358+ {
359+ set_error(d, "Invalid attribute");
360+ return start_length(0, 0);
361+ }
344362}
345363
346364
347365
348366// seen a tag name, now looking for attributes terminated by >
349367// puts the attributes it finds in the attribute buffer
350−str static inline parse_attributes(document* d)
368+static inline str parse_attributes(document* d)
351369{
352− str res;
353− res.start = d->attrs.used;
354− for (int i = 0; ; i++)
370+ int start = d->attrs.used;
371+ while (d->error_message == NULL)
355372 {
373+ trim(d);
356374 str name = parse_name(d);
357375 if (name.length == 0) break;
358376 attr_alloc(&d->attrs, 1);
@@ -360,34 +378,39 @@ str static inline parse_attributes(document* d)
360378 d->attrs.attrs[d->attrs.used].value = parse_attrval(d);
361379 d->attrs.used++;
362380 }
363− res.length = d->attrs.used - res.start;
364− return res;
381+ return start_end(start, d->attrs.used);
365382}
366383
367−str parse_content(document* d);
384+static str parse_content(document* d);
368385
369386
370387// Add a new entry into tag, am at a '<'
371−void static inline parse_tag(document* d)
388+static inline void parse_tag(document* d)
372389{
373390 node_alloc(&d->nodes, 1);
374391 d->nodes.used_back++;
375− int me = d->nodes.size - d->nodes.used_back;
392+ node* me = &d->nodes.nodes[d->nodes.size - d->nodes.used_back];
376393
377− d->nodes.nodes[me].outer.start = doc_position(d);
394+ me->outer.start = doc_position(d);
378395 char c = get(d);
379396 assert(c == '<');
380397 if (peek(d) == '?') skip(d, 1);
381− d->nodes.nodes[me].name = parse_name(d);
382− 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;
383406
384− c = get(d);
385− if ((c == '/' || c == '?') && peek(d) == '>')
407+ c = peek(d);
408+ if ((c == '/' || c == '?') && peek_at(d, 1) == '>')
386409 {
387− skip(d, 1);
388− d->nodes.nodes[me].nodes = start_length(0, 0);
389− d->nodes.nodes[me].outer.length = start_end(d->nodes.nodes[me].outer.start, doc_position(d)).length;
390− 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);
391414 return;
392415 }
393416 else if (c != '>')
@@ -395,23 +418,28 @@ void static inline parse_tag(document* d)
395418 set_error(d, "Gunk at the end of the tag");
396419 return;
397420 }
398− d->nodes.nodes[me].inner.start = doc_position(d);
399− d->nodes.nodes[me].nodes = parse_content(d);
400− 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;
401430
402431 if (d->error_message != NULL) return;
403− if (peek(d) == '<' && peekAt(d, 1) == '/')
432+ if (peek(d) == '<' && peek_at(d, 1) == '/')
404433 {
405434 skip(d, 2);
406− trim(d);
407− if (d->end - d->cursor >= d->nodes.nodes[me].name.length &&
408− 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)
409437 {
410− skip(d, d->nodes.nodes[me].name.length);
438+ skip(d, me->name.length);
411439 trim(d);
412440 if (get(d) == '>')
413441 {
414− 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;
415443 return;
416444 }
417445 }
@@ -423,7 +451,7 @@ void static inline parse_tag(document* d)
423451
424452// Parser until </, return the index of your node children
425453// Not inline as it is recursive
426−str parse_content(document* d)
454+static str parse_content(document* d)
427455{
428456 int before = d->nodes.used_back;
429457 while (d->error_message == NULL)
@@ -433,11 +461,33 @@ str parse_content(document* d)
433461 {
434462 break;
435463 }
436− if (peekAt(d, 1) == '/')
464+ else if (peek_at(d, 1) == '/')
437465 {
438466 // have found a </
439467 break;
440468 }
469+ else if (peek_at(d, 1) == '!' && peek_at(d, 2) == '-' && peek_at(d, 3) == '-')
470+ {
471+ skip(d, 3);
472+ // you can't reuse the two '-' characters for the closing as well
473+ if (peek_at(d, 0) == '\0' || peek_at(d, 1) == '\0')
474+ {
475+ set_error(d, "Didn't get a closing comment");
476+ return start_end(0, 0);
477+ }
478+ skip(d, 2);
479+ for (;;)
480+ {
481+ if (!find(d, '>'))
482+ {
483+ set_error(d, "Didn't get a closing comment");
484+ return start_end(0, 0);
485+ }
486+ skip(d, 1);
487+ if (peek_at(d, -3) == '-' && peek_at(d, -2) == '-')
488+ break;
489+ }
490+ }
441491 else
442492 {
443493 parse_tag(d);
@@ -453,20 +503,23 @@ str parse_content(document* d)
453503 return res;
454504}
455505
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.
456508typedef struct
457509{
458510 document document;
459511 attr attrs[1000];
460− node nodes[1000];
512+ node nodes[500];
461513} buffer;
462514
463−document* document_parse(char* s, int slen)
515+document* hexml_document_parse(const char* s, int slen)
464516{
465− if (slen == -1) slen = strlen(s);
517+ if (slen == -1) slen = (int) strlen(s);
466518 assert(s[slen] == 0);
467519 init_parse_table();
468520
469521 buffer* buf = malloc(sizeof(buffer));
522+ assert(buf);
470523 document* d = &buf->document;
471524 d->body = s;
472525 d->cursor = s;
@@ -476,7 +529,7 @@ document* document_parse(char* s, int slen)
476529 d->attrs.used = 0;
477530 d->attrs.attrs = buf->attrs;
478531 d->attrs.alloc = NULL;
479− d->nodes.size = 1000;
532+ d->nodes.size = 500;
480533 d->nodes.used_back = 0;
481534 d->nodes.used_front = 1;
482535 d->nodes.nodes = buf->nodes;
@@ -486,11 +539,13 @@ document* document_parse(char* s, int slen)
486539 d->nodes.nodes[0].outer = start_length(0, slen);
487540 d->nodes.nodes[0].inner = start_length(0, slen);
488541 d->nodes.nodes[0].attrs = start_length(0, 0);
489− d->nodes.nodes[0].nodes = parse_content(d);
490542
491− if (d->cursor < d->end && d->error_message == NULL)
492− {
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)
493549 set_error(d, "Trailing junk at the end of the document");
494− }
495550 return d;
496551}