sd_ass: allow get_text to return more than 500 bytes

rcombs committed 2 years ago

Commit aa0a9ce1 parent e3fd2444 files · +56 −58Browse files

Unified · Split

4 files

sub/filter_jsre.c +1 −1
sub/filter_regex.c +1 −1
sub/sd.h +5 −3
sub/sd_ass.c +49 −53

@@ -115,7 +115,7 @@ static struct demux_packet jsre_filter(struct sd_filter ft, · 114 unmodified lines
115	bool drop = false;	115	bool drop = false;
116		116
117	if (ft->opts->rf_plain)	117	if (ft->opts->rf_plain)
118	sd_ass_to_plaintext(text, strlen(text), text);	118	sd_ass_to_plaintext(&text, text);
119		119
120	for (int n = 0; n < p->num_regexes; n++) {	120	for (int n = 0; n < p->num_regexes; n++) {
121	int found, err = p_regexec(p->J, n, text, &found);	121	int found, err = p_regexec(p->J, n, text, &found);

sub/filter_regex.c+1−1

@@ -64,7 +64,7 @@ static struct demux_packet rf_filter(struct sd_filter ft, · 63 unmodified lines
64	bool drop = false;	64	bool drop = false;
65		65
66	if (ft->opts->rf_plain)	66	if (ft->opts->rf_plain)
67	sd_ass_to_plaintext(text, strlen(text), text);	67	sd_ass_to_plaintext(&text, text);
68		68
69	for (int n = 0; n < p->num_regexes; n++) {	69	for (int n = 0; n < p->num_regexes; n++) {
70	int err = regexec(&p->regexes[n], text, 0, NULL, 0);	70	int err = regexec(&p->regexes[n], text, 0, NULL, 0);

sub/sd.h+5−3

 bstr sd_ass_pkt_text(struct sd_filter *ft, struct demux_packet *pkt, int offset);
 // convert \0-terminated "Text" (ass) content to plaintext, possibly in-place.
-// result.start is out, result.len is MIN(out_siz, strlen(in)) or smaller.
+// result.start is *out, result.len is strlen(in) or smaller.
-// if there's room: out[result.len] is set to \0. out == in is allowed.
+// (*out)[result.len] is always set to \0. *out == in is allowed.
-bstr sd_ass_to_plaintext(char *out, size_t out_siz, const char *in);
+// *out must be a talloc-allocated buffer or NULL, and will be reallocated if needed.
+// *out will not be reallocated if *out == in.
+bstr sd_ass_to_plaintext(char **out, const char *in);
 #endif

sub/sd_ass.c+49−53

     bool clear_once;
     struct mp_ass_packer *packer;
     struct sub_bitmap_copy_cache *copy_cache;
-    char last_text[500];
+    bstr last_text;
     struct mp_image_params video_params;
     struct mp_image_params last_params;
     struct mp_osd_res osd;
     return res;
 }
-struct buf {
+#define MAX_BUF_SIZE 1024 * 1024
-    char *start;
+#define MIN_EXPAND_SIZE 4096
-    int size;
-    int len;
-};
-static void append(struct buf *b, char c)
+static void append(bstr *b, char c)
 {
-    if (b->len < b->size) {
+    bstr_xappend(NULL, b, (bstr){&c, 1});
-        b->start[b->len] = c;
-        b->len++;
-    }
 }
-static void ass_to_plaintext(struct buf *b, const char *in)
+static void ass_to_plaintext(bstr *b, const char *in)
 {
-    bool in_tag = false;
     const char *open_tag_pos = NULL;
     bool in_drawing = false;
     while (*in) {
-        if (in_tag) {
+        if (open_tag_pos) {
             if (in[0] == '}') {
                 in += 1;
-                in_tag = false;
+                open_tag_pos = NULL;
             } else if (in[0] == '\\' && in[1] == 'p' && in[2] != 'o') {
                 in += 2;
                 // Skip text between \pN and \p0 tags. A \p without a number
             } else if (in[0] == '{') {
                 open_tag_pos = in;
                 in += 1;
-                in_tag = true;
             } else {
                 if (!in_drawing)
                     append(b, in[0]);
         }
     }
     // A '{' without a closing '}' is always visible.
-    if (in_tag) {
+    if (open_tag_pos) {
-        while (*open_tag_pos)
+        bstr_xappend(NULL, b, bstr0(open_tag_pos));
-            append(b, *open_tag_pos++);
     }
 }
-// Empty string counts as whitespace. Reads s[len-1] even if there are \0s.
+// Empty string counts as whitespace.
-static bool is_whitespace_only(char *s, int len)
+static bool is_whitespace_only(bstr b)
 {
-    for (int n = 0; n < len; n++) {
+    for (int n = 0; n < b.len; n++) {
-        if (s[n] != ' ' && s[n] != '\t')
+        if (b.start[n] != ' ' && b.start[n] != '\t')
             return false;
     }
     return true;
 }
-static char *get_text_buf(struct sd *sd, double pts, enum sd_text_type type)
+static bstr get_text_buf(struct sd *sd, double pts, enum sd_text_type type)
 {
     struct sd_ass_priv *ctx = sd->priv;
     ASS_Track *track = ctx->ass_track;
     if (pts == MP_NOPTS_VALUE)
-        return NULL;
+        return (bstr){0};
     long long ipts = find_timestamp(sd, pts);
-    struct buf b = {ctx->last_text, sizeof(ctx->last_text) - 1};
+    bstr *b = &ctx->last_text;
+    if (!b->start)
+        b->start = talloc_size(ctx, 4096);
+    b->len = 0;
     for (int i = 0; i < track->n_events; ++i) {
         ASS_Event *event = track->events + i;
         if (ipts >= event->Start && ipts < event->Start + event->Duration) {
             if (event->Text) {
-                int start = b.len;
+                int start = b->len;
                 if (type == SD_TEXT_TYPE_PLAIN) {
-                    ass_to_plaintext(&b, event->Text);
+                    ass_to_plaintext(b, event->Text);
                 } else {
-                    char *t = event->Text;
+                    bstr_xappend(NULL, b, bstr0(event->Text));
-                    while (*t)
-                        append(&b, *t++);
                 }
-                if (is_whitespace_only(&b.start[start], b.len - start)) {
+                if (is_whitespace_only(bstr_cut(*b, start))) {
-                    b.len = start;
+                    b->len = start;
                 } else {
-                    append(&b, '\n');
+                    append(b, '\n');
                 }
             }
         }
     }
-    b.start[b.len] = '\0';
+    bstr_eatend(b, (bstr)bstr0_lit("\n"));
-    if (b.len > 0 && b.start[b.len - 1] == '\n')
-        b.start[b.len - 1] = '\0';
-    return ctx->last_text;
+    return *b;
 }
 static char *get_text(struct sd *sd, double pts, enum sd_text_type type)
 {
-    return talloc_strdup(NULL, get_text_buf(sd, pts, type));
+    return bstrto0(NULL, get_text_buf(sd, pts, type));
 }
 static struct sd_times get_times(struct sd *sd, double pts)
     ass_flush_events(track);
-    char *text = get_text_buf(sd, pts, SD_TEXT_TYPE_PLAIN);
+    bstr text = get_text_buf(sd, pts, SD_TEXT_TYPE_PLAIN);
-    if (!text)
+    if (!text.len)
         return;
     bstr dst = {0};
-    while (*text) {
+    while (text.len) {
-        if (*text == '{')
+        if (*text.start == '{') {
+            bstr_xappend(NULL, &dst, bstr0("\\{"));
+            text = bstr_cut(text, 1);
+        } else if (*text.start == '\\') {
             bstr_xappend(NULL, &dst, bstr0("\\"));
-        bstr_xappend(NULL, &dst, (bstr){text, 1});
+            // Break ASS escapes with U+2060 WORD JOINER
-        // Break ASS escapes with U+2060 WORD JOINER
-        if (*text == '\\')
             mp_append_utf8_bstr(NULL, &dst, 0x2060);
-        text++;
+            text = bstr_cut(text, 1);
+        }
+        int i = bstrcspn(text, "{\\");
+        bstr_xappend(NULL, &dst, (bstr){text.start, i});
+        text = bstr_cut(text, i);
     }
     if (!dst.start)
     return txt;
 }
-bstr sd_ass_to_plaintext(char *out, size_t out_siz, const char *in)
+bstr sd_ass_to_plaintext(char **out, const char *in)
 {
-    struct buf b = {out, out_siz, 0};
+    bstr b = {*out};
     ass_to_plaintext(&b, in);
-    if (b.len < out_siz)
+    *out = b.start;
-        out[b.len] = 0;
+    return b;
-    return (bstr){out, b.len};
 }

sd_ass: allow get_text to return more than 500 bytes

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code