aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorDaniel Eklöf <daniel@ekloef.se>2025-01-30 08:51:50 +0000
committerDaniel Eklöf <daniel@ekloef.se>2025-02-05 12:35:16 +0000
commit859b4c8921f5cde8cfc1f358e865ba417fe68a12 (patch)
treef097e3fb2dffeb8c40701e91e1d665a05cc030f6
parenturl-mode: wip: convert to regex matching for auto-detection (diff)
url-mode: wip: more work on regex matching
Remove the old auto-detection and instead use the regex matches.
-rw-r--r--url-mode.c361
1 files changed, 75 insertions, 286 deletions
diff --git a/url-mode.c b/url-mode.c
index 0ed79a99..18085ab5 100644
--- a/url-mode.c
+++ b/url-mode.c
@@ -292,21 +292,31 @@ urls_input(struct seat *seat, struct terminal *term,
}
}
-struct charmap {
- const struct row *row;
- int col;
-};
-
struct vline {
char *utf8;
- size_t len;
- size_t sz;
- struct charmap *map;
+ size_t len; /* Length of utf8[] */
+ size_t sz; /* utf8[] allocated size */
+ struct coord *map; /* Maps utf8[ofs] to grid coordinates */
};
static void
regex_detected(const struct terminal *term, enum url_action action, url_list_t *urls)
{
+ /*
+ * Use regcomp()+regexec() to find patterns.
+ *
+ * Since we can't feed regexec() one character at a time, and
+ * since it doesn't accept wide characters, we need to build utf8
+ * strings.
+ *
+ * Each string represents a logical line (i.e. handle line-wrap).
+ * To be able to map regex matches back to the grid, we store the
+ * grid coordinates of *each* character, in the line struct as
+ * well. This is offset based; utf8[ofs] has its grid coordinates
+ * in map[ofs.
+ */
+
+ /* There is *at most* term->rows logical lines */
struct vline vlines[term->rows];
size_t vline_idx = 0;
@@ -315,14 +325,15 @@ regex_detected(const struct terminal *term, enum url_action action, url_list_t *
mbstate_t ps = {0};
- for (int row_no = 0; row_no < term->rows; row_no++) {
- const struct row *row = grid_row_in_view(term->grid, row_no);
+ for (int r = 0; r < term->rows; r++) {
+ const struct row *row = grid_row_in_view(term->grid, r);
for (int c = 0; c < term->cols; c++) {
const struct cell *cell = &row->cells[c];
const char32_t *wc = &cell->wc;
size_t wc_count = 1;
+ /* Expand combining characters */
if (wc[0] >= CELL_COMB_CHARS_LO && wc[0] <= CELL_COMB_CHARS_HI) {
const struct composed *composed =
composed_lookup(term->composed, wc[0] - CELL_COMB_CHARS_LO);
@@ -332,26 +343,26 @@ regex_detected(const struct terminal *term, enum url_action action, url_list_t *
wc_count = composed->count;
}
+ /* Convert wide character to utf8 */
for (size_t i = 0; i < wc_count; i++) {
char buf[16];
size_t char_len = c32rtomb(buf, wc[i], &ps);
- if (char_len == (size_t)-1)
- continue;
- if (vline->len == 0 && char_len == 1 && buf[0] == 0)
+ if (char_len == (size_t)-1)
continue;
for (size_t j = 0; j < char_len; j++) {
- if (vline->len + char_len > vline->sz) {
- /* TODO: grow dynamically */
- size_t new_count = (vline->len + char_len) * 2;
- vline->utf8 = xreallocarray(vline->utf8, new_count, 1);
- vline->map = xreallocarray(vline->map, new_count, sizeof(vline->map[0]));
+ const size_t requires_size = vline->len + char_len;
+
+ if (requires_size > vline->sz) {
+ const size_t new_size = requires_size * 2;
+ vline->utf8 = xreallocarray(vline->utf8, new_size, 1);
+ vline->map = xreallocarray(vline->map, new_size, sizeof(vline->map[0]));
+ vline->sz = new_size;
}
vline->utf8[vline->len + j] = buf[j];
- vline->map[vline->len + j].col = c;
- vline->map[vline->len + j].row = row;
+ vline->map[vline->len + j] = (struct coord){c, term->grid->view + r};
}
vline->len += char_len;
@@ -371,9 +382,9 @@ regex_detected(const struct terminal *term, enum url_action action, url_list_t *
// https://gist.github.com/gruber/249502
regex_t preg;
- const char *foo =
+ const char *regex_string =
"("
- "[a-z][[:alpha:]-]+:" // protocol
+ "[a-z][[:alnum:]-]+:" // protocol
"("
"/{1,3}|[a-z0-9%]" // slashes (what's the OR part for?)
")"
@@ -390,291 +401,70 @@ regex_detected(const struct terminal *term, enum url_action action, url_list_t *
"("
"\\(([^[:space:]()<>]+|(\\([^[:space:]()<>]+\\)))*\\)"
"|"
- // TODO: figure out how to add \\] to the expression below...
- "[^[:space:]`!()\\[{};:'\".,<>?«»“”‘’]"
+ "[^]\\[[:space:]`!(){};:'\".,<>?«»“”‘’]"
")"
;
- LOG_ERR("foo=%s", foo);
-
- int r = regcomp(&preg, foo, REG_EXTENDED);
+ int r = regcomp(&preg, regex_string, REG_EXTENDED);
if (r != 0) {
char err_buf[1024];
regerror(r, &preg, err_buf, sizeof(err_buf));
- LOG_ERR("regcomp: %s", err_buf);
- } else {
- size_t i = 0;
- while (true) {
- const struct vline *v = &vlines[i++];
- if (v->utf8 == NULL)
- break;
-
-
- regmatch_t matches[preg.re_nsub + 1];
- r = regexec(&preg, v->utf8, preg.re_nsub + 1, matches, 0);
-
- if (r == REG_NOMATCH)
- continue;
+ LOG_ERR("failed to compile regular expression: %s", err_buf);
- size_t mlen = matches[0].rm_eo - matches[0].rm_so;
- LOG_WARN("MATCH at %d: %.*s (%zu)", matches[0].rm_so, (int)mlen, &v->utf8[matches[0].rm_so], mlen);
+ for (size_t i = 0; i < ALEN(vlines); i++) {
+ const struct vline *v = &vlines[i];
+ free(v->utf8);
+ free(v->map);
}
- regfree(&preg);
- }
-
- size_t i = 0;
- while (true) {
- const struct vline *v = &vlines[i++];
- if (v->utf8 == NULL)
- break;
-
- LOG_WARN("%.*s", (int)v->len, v->utf8);
- free(v->utf8);
- free(v->map);
- }
-}
-
-static int
-c32cmp_single(const void *_a, const void *_b)
-{
- const char32_t *a = _a;
- const char32_t *b = _b;
- return *a - *b;
-}
-static void
-auto_detected(const struct terminal *term, enum url_action action,
- url_list_t *urls)
-{
- const struct config *conf = term->conf;
-
- const char32_t *uri_characters = conf->url.uri_characters;
- if (uri_characters == NULL)
return;
+ }
- const size_t uri_characters_count = c32len(uri_characters);
- if (uri_characters_count == 0)
- return;
-
- size_t max_prot_len = conf->url.max_prot_len;
- char32_t proto_chars[max_prot_len];
- struct coord proto_start[max_prot_len];
- size_t proto_char_count = 0;
-
- enum {
- STATE_PROTOCOL,
- STATE_URL,
- } state = STATE_PROTOCOL;
-
- struct coord start = {-1, -1};
- char32_t url[term->cols * term->rows + 1];
- size_t len = 0;
-
- ssize_t parenthesis = 0;
- ssize_t brackets = 0;
- ssize_t ltgts = 0;
-
- for (int r = 0; r < term->rows; r++) {
- const struct row *row = grid_row_in_view(term->grid, r);
-
- for (int c = 0; c < term->cols; c++) {
- const struct cell *cell = &row->cells[c];
-
- if (cell->wc >= CELL_SPACER)
- continue;
-
- const char32_t *wcs = NULL;
- size_t wc_count = 0;
-
- if (cell->wc >= CELL_COMB_CHARS_LO && cell->wc <= CELL_COMB_CHARS_HI) {
- struct composed *composed =
- composed_lookup(term->composed, cell->wc - CELL_COMB_CHARS_LO);
- wcs = composed->chars;
- wc_count = composed->count;
- } else {
- wcs = &cell->wc;
- wc_count = 1;
- }
-
- for (size_t w_idx = 0; w_idx < wc_count; w_idx++) {
- char32_t wc = wcs[w_idx];
-
- switch (state) {
- case STATE_PROTOCOL:
- for (size_t i = 0; i < max_prot_len - 1; i++) {
- proto_chars[i] = proto_chars[i + 1];
- proto_start[i] = proto_start[i + 1];
- }
-
- if (proto_char_count >= max_prot_len)
- proto_char_count = max_prot_len - 1;
-
- proto_chars[max_prot_len - 1] = wc;
- proto_start[max_prot_len - 1] = (struct coord){c, r};
- proto_char_count++;
-
- for (size_t i = 0; i < conf->url.prot_count; i++) {
- size_t prot_len = c32len(conf->url.protocols[i]);
-
- if (proto_char_count < prot_len)
- continue;
-
- const char32_t *proto =
- &proto_chars[max_prot_len - prot_len];
-
- if (c32ncasecmp(conf->url.protocols[i], proto, prot_len) ==
- 0) {
- state = STATE_URL;
- start = proto_start[max_prot_len - prot_len];
-
- c32ncpy(url, proto, prot_len);
- len = prot_len;
-
- parenthesis = brackets = ltgts = 0;
- break;
- }
- }
- break;
-
- case STATE_URL: {
- const char32_t *match =
- bsearch(&wc, uri_characters, uri_characters_count,
- sizeof(uri_characters[0]), &c32cmp_single);
-
- bool emit_url = false;
-
- if (match == NULL) {
- /*
- * Character is not a valid URI character. Emit
- * the URL we've collected so far, *without*
- * including _this_ character.
- */
- emit_url = true;
- } else {
- xassert(*match == wc);
-
- switch (wc) {
- default:
- url[len++] = wc;
- break;
-
- case U'(':
- parenthesis++;
- url[len++] = wc;
- break;
-
- case U'[':
- brackets++;
- url[len++] = wc;
- break;
-
- case U'<':
- ltgts++;
- url[len++] = wc;
- break;
-
- case U')':
- if (--parenthesis < 0)
- emit_url = true;
- else
- url[len++] = wc;
- break;
-
- case U']':
- if (--brackets < 0)
- emit_url = true;
- else
- url[len++] = wc;
- break;
-
- case U'>':
- if (--ltgts < 0)
- emit_url = true;
- else
- url[len++] = wc;
- break;
- }
- }
-
- if (c >= term->cols - 1 && row->linebreak) {
- /*
- * Endpoint is inclusive, and we'll be subtracting
- * 1 from the column when emitting the URL.
- */
- c++;
- emit_url = true;
- }
-
- if (emit_url) {
- struct coord end = {c, r};
+ for (size_t i = 0; i < ALEN(vlines); i++) {
+ const struct vline *v = &vlines[i];
+ if (v->utf8 == NULL)
+ continue;;
- if (--end.col < 0) {
- end.row--;
- end.col = term->cols - 1;
- }
+ const char *search_string = v->utf8;
+ while (true) {
- /* Heuristic to remove trailing characters that
- * are valid URL characters, but typically not at
- * the end of the URL */
- bool done = false;
- do {
- switch (url[len - 1]) {
- case U'.':
- case U',':
- case U':':
- case U';':
- case U'?':
- case U'!':
- case U'"':
- case U'\'':
- case U'%':
- len--;
- end.col--;
- if (end.col < 0) {
- end.row--;
- end.col = term->cols - 1;
- }
- break;
+ regmatch_t matches[preg.re_nsub + 1];
+ r = regexec(&preg, search_string, preg.re_nsub + 1, matches, 0);
- default:
- done = true;
- break;
- }
- } while (!done);
+ if (r == REG_NOMATCH)
+ break;
- url[len] = U'\0';
+ const size_t mlen = matches[0].rm_eo - matches[0].rm_so;
+ const size_t start = &search_string[matches[0].rm_so] - v->utf8;
+ const size_t end = start + mlen;
- start.row += term->grid->view;
- end.row += term->grid->view;
+ LOG_DBG(
+ "MATCH at %d: %.*s (%zu) row/col = %dx%d",
+ matches[0].rm_so, (int)mlen, &search_string[matches[0].rm_so],
+ mlen, v->map[start].row, v->map[start].col);
- char *url_utf8 = ac32tombs(url);
- if (url_utf8 != NULL) {
- tll_push_back(
- *urls,
- ((struct url){.id = (uint64_t)rand() << 32 | rand(),
- .url = url_utf8,
- .range =
- {
- .start = start,
- .end = end,
- },
- .action = action,
- .osc8 = false}));
- }
+ tll_push_back(
+ *urls,
+ ((struct url){
+ .id = (uint64_t)rand() << 32 | rand(),
+ .url = xstrndup(&v->utf8[start], mlen),
+ .range = {
+ .start = v->map[start],
+ .end = v->map[end - 1], /* Inclusive */
+ },
+ .action = action,
+ .osc8 = false}));
- state = STATE_PROTOCOL;
- len = 0;
- parenthesis = brackets = ltgts = 0;
- }
- break;
- }
- }
- }
+ search_string += matches[0].rm_eo;
}
+
+ free(v->utf8);
+ free(v->map);
}
-}
+ regfree(&preg);
+}
static void
osc8_uris(const struct terminal *term, enum url_action action, url_list_t *urls)
{
@@ -777,7 +567,6 @@ urls_collect(const struct terminal *term, enum url_action action, url_list_t *ur
{
xassert(tll_length(term->urls) == 0);
osc8_uris(term, action, urls);
- auto_detected(term, action, urls);
regex_detected(term, action, urls);
remove_overlapping(urls, term->grid->num_cols);
}