Add thumbnail-count option

third-party/libscan/libscan/comic/comic.c

@@ -12,7 +12,7 @@ void parse_comic(scan_comic_ctx_t *ctx, vfile_t *f, document_t *doc) {
     struct archive_entry *entry = NULL;
     arc_data_t arc_data;
 
-    if (ctx->tn_size <= 0) {
+    if (!ctx->enable_tn) {
         return;
     }
 
@@ -44,7 +44,20 @@ void parse_comic(scan_comic_ctx_t *ctx, vfile_t *f, document_t *doc) {
                     break;
                 }
 
-                ret = store_image_thumbnail((scan_media_ctx_t *) ctx, buf, entry_size, doc, file_path);
+                scan_media_ctx_t media_ctx = {
+                        .tn_count = ctx->enable_tn ? 1 : 0,
+                        .tn_size = ctx->tn_size,
+                        .tn_qscale = ctx->tn_qscale,
+                        .tesseract_lang = NULL,
+                        .tesseract_path = NULL,
+                        .read_subtitles = FALSE,
+                        .max_media_buffer = 0,
+                        .log = ctx->log,
+                        .logf = ctx->logf,
+                        .store = ctx->store,
+                };
+
+                ret = store_image_thumbnail(&media_ctx, buf, entry_size, doc, file_path);
                 free(buf);
 
                 if (ret == TRUE) {

third-party/libscan/libscan/comic/comic.h

@@ -9,6 +9,7 @@ typedef struct {
     logf_callback_t logf;
     store_callback_t store;
 
+    int enable_tn;
     int tn_size;
     float tn_qscale;
 

third-party/libscan/libscan/ebook/ebook.c

@@ -155,7 +155,7 @@ int render_cover(scan_ebook_ctx_t *ctx, fz_context *fzctx, document_t *doc, fz_d
     av_init_packet(&jpeg_packet);
     avcodec_receive_packet(jpeg_encoder, &jpeg_packet);
 
-    APPEND_TN_META(doc, pixmap->w, pixmap->h)
+    APPEND_LONG_META(doc, MetaThumbnail, 1)
     ctx->store((char *) doc->path_md5, sizeof(doc->path_md5), (char *) jpeg_packet.data, jpeg_packet.size);
 
     free(samples);
@@ -283,7 +283,7 @@ parse_ebook_mem(scan_ebook_ctx_t *ctx, void *buf, size_t buf_len, const char *mi
 
     APPEND_LONG_META(doc, MetaPages, page_count)
 
-    if (ctx->tn_size > 0) {
+    if (ctx->enable_tn) {
         if (render_cover(ctx, fzctx, doc, fzdoc) == FALSE) {
             fz_drop_stream(fzctx, stream);
             fz_drop_document(fzctx, fzdoc);
@@ -404,7 +404,7 @@ void parse_epub_fast(scan_ebook_ctx_t *ctx, vfile_t *f, document_t *doc) {
 
     text_buffer_t content_buffer = text_buffer_create(ctx->content_size);
 
-    if (ctx->tn_size <= 0) {
+    if (!ctx->enable_tn) {
         return;
     }
 

third-party/libscan/libscan/ebook/ebook.h

@@ -6,6 +6,7 @@
 typedef struct {
     long content_size;
     int tn_size;
+    int enable_tn;
     const char *tesseract_lang;
     const char *tesseract_path;
     pthread_mutex_t mupdf_mutex;

third-party/libscan/libscan/font/font.c

@@ -176,7 +176,7 @@ void parse_font(scan_font_ctx_t *ctx, vfile_t *f, document_t *doc) {
     strcpy(meta_name->str_val, font_name);
     APPEND_META(doc, meta_name)
 
-    if (ctx->enable_tn == TRUE) {
+    if (!ctx->enable_tn) {
         FT_Done_Face(face);
         free(buf);
         return;
@@ -231,7 +231,7 @@ void parse_font(scan_font_ctx_t *ctx, vfile_t *f, document_t *doc) {
     dyn_buffer_t bmp_data = dyn_buffer_create();
     bmp_format(&bmp_data, dimensions, bitmap);
 
-    APPEND_TN_META(doc, dimensions.width, dimensions.height)
+    APPEND_LONG_META(doc, MetaThumbnail, 1)
     ctx->store((char *) doc->path_md5, sizeof(doc->path_md5), (char *) bmp_data.buf, bmp_data.cur);
 
     dyn_buffer_destroy(&bmp_data);

third-party/libscan/libscan/macros.h

@@ -35,11 +35,6 @@
     meta_long->long_val = value; \
     APPEND_META(doc, meta_long)}
 
-#define APPEND_TN_META(doc, width, height) \
-    {meta_line_t *meta_str = malloc(sizeof(meta_line_t) + 4 + 1 + 4); \
-    meta_str->key = MetaThumbnail; \
-    sprintf(meta_str->str_val, "%04d,%04d", width, height); \
-    APPEND_META(doc, meta_str)}
 
 #define APPEND_META(doc, meta) \
     meta->next = NULL;\

third-party/libscan/libscan/media/media.c

@@ -6,7 +6,6 @@
 #define AVIO_BUF_SIZE 8192
 #define IS_VIDEO(fmt) ((fmt)->iformat->name && strcmp((fmt)->iformat->name, "image2") != 0)
 
-#define STREAM_IS_IMAGE (stream->nb_frames <= 1)
 
 #define STORE_AS_IS ((void*)-1)
 
@@ -398,6 +397,110 @@ void ocr_image(scan_media_ctx_t *ctx, document_t *doc, const AVCodecContext *dec
     av_frame_free(&rgb_frame);
 }
 
+#define SAVE_THUMBNAIL_OK 0
+#define SAVE_THUMBNAIL_SKIPPED 1
+#define SAVE_THUMBNAIL_FAILED 2
+
+int decode_frame_and_save_thumbnail(scan_media_ctx_t *ctx, AVFormatContext *pFormatCtx, AVCodecContext *decoder,
+                                    AVStream *stream, int video_stream, document_t *doc, double seek_ratio,
+                                    int thumbnail_index) {
+
+    if (IS_VIDEO(pFormatCtx) && stream->codecpar->codec_id != AV_CODEC_ID_GIF) {
+        int seek_ok = FALSE;
+
+        double target_timestamp = (double) pFormatCtx->duration * seek_ratio;
+        long ts = (long) target_timestamp;
+
+        int seek_ret = avformat_seek_file(
+                // Allow +- 1s
+                pFormatCtx, -1, ts - AV_TIME_BASE, ts, ts + AV_TIME_BASE,
+                0
+        );
+
+        if (seek_ret == 0) {
+            seek_ok = TRUE;
+        } else {
+            CTX_LOG_DEBUGF(
+                    doc->filepath,
+                    "(media.c) Could not seek media file: %s", av_err2str(seek_ret)
+            )
+        }
+
+        if (seek_ok == FALSE && thumbnail_index != 0) {
+            CTX_LOG_WARNING(doc->filepath, "(media.c) Could not seek media file. Can't generate additional thumbnails.")
+            return SAVE_THUMBNAIL_FAILED;
+        }
+    }
+
+    frame_and_packet_t *frame_and_packet = read_frame(ctx, pFormatCtx, decoder, video_stream, doc);
+    if (frame_and_packet == NULL) {
+        return SAVE_THUMBNAIL_FAILED;
+    }
+
+    if (ctx->tesseract_lang != NULL && IS_VIDEO(pFormatCtx)) {
+        ocr_image(ctx, doc, decoder, frame_and_packet->frame);
+    }
+
+    // NOTE: OCR'd content takes precedence over exif image description
+    if (thumbnail_index == 0) {
+        append_video_meta(ctx, pFormatCtx, frame_and_packet->frame, doc, IS_VIDEO(pFormatCtx));
+    }
+
+    // Scale frame
+    AVFrame *scaled_frame = scale_frame(decoder, frame_and_packet->frame, ctx->tn_size);
+
+    if (scaled_frame == NULL) {
+        frame_and_packet_free(frame_and_packet);
+        return SAVE_THUMBNAIL_FAILED;
+    }
+
+    int return_value;
+
+    if (scaled_frame == STORE_AS_IS) {
+        return_value = SAVE_THUMBNAIL_OK;
+
+        ctx->store((char *) doc->path_md5, sizeof(doc->path_md5), (char *) frame_and_packet->packet->data,
+                   frame_and_packet->packet->size);
+    } else {
+        // Encode frame to jpeg
+        AVCodecContext *jpeg_encoder = alloc_jpeg_encoder(scaled_frame->width, scaled_frame->height,
+                                                          ctx->tn_qscale);
+        avcodec_send_frame(jpeg_encoder, scaled_frame);
+
+        AVPacket jpeg_packet;
+        av_init_packet(&jpeg_packet);
+        avcodec_receive_packet(jpeg_encoder, &jpeg_packet);
+
+        // Save thumbnail
+        if (thumbnail_index == 0) {
+            ctx->store((char *) doc->path_md5, sizeof(doc->path_md5), (char *) jpeg_packet.data, jpeg_packet.size);
+            return_value = SAVE_THUMBNAIL_OK;
+
+        } else if (thumbnail_index > 1) {
+            return_value = SAVE_THUMBNAIL_OK;
+            // TO FIX: the 2nd rendered frame is always broken, just skip it until
+            //  I figure out a better fix.
+            thumbnail_index -= 1;
+
+            char tn_key[sizeof(doc->path_md5) + sizeof(int)];
+            memcpy(tn_key, doc->path_md5, sizeof(doc->path_md5));
+            memcpy(tn_key + sizeof(doc->path_md5), &thumbnail_index, sizeof(thumbnail_index));
+
+            ctx->store((char *) tn_key, sizeof(tn_key), (char *) jpeg_packet.data, jpeg_packet.size);
+        } else {
+            return_value = SAVE_THUMBNAIL_SKIPPED;
+        }
+
+        avcodec_free_context(&jpeg_encoder);
+        av_packet_unref(&jpeg_packet);
+        av_free(*scaled_frame->data);
+        av_frame_free(&scaled_frame);
+    }
+
+    frame_and_packet_free(frame_and_packet);
+    return return_value;
+}
+
 void parse_media_format_ctx(scan_media_ctx_t *ctx, AVFormatContext *pFormatCtx, document_t *doc) {
 
     int video_stream = -1;
@@ -458,7 +561,7 @@ void parse_media_format_ctx(scan_media_ctx_t *ctx, AVFormatContext *pFormatCtx,
         append_audio_meta(pFormatCtx, doc);
     }
 
-    if (video_stream != -1 && ctx->tn_size > 0) {
+    if (video_stream != -1 && ctx->tn_count > 0) {
         AVStream *stream = pFormatCtx->streams[video_stream];
 
         if (stream->codecpar->width <= MIN_SIZE || stream->codecpar->height <= MIN_SIZE) {
@@ -473,69 +576,38 @@ void parse_media_format_ctx(scan_media_ctx_t *ctx, AVFormatContext *pFormatCtx,
         avcodec_parameters_to_context(decoder, stream->codecpar);
         avcodec_open2(decoder, video_codec, NULL);
 
-        //Seek
-        if (!STREAM_IS_IMAGE && stream->codecpar->codec_id != AV_CODEC_ID_GIF) {
-            int seek_ret;
-            for (int i = 20; i >= 0; i--) {
-                seek_ret = av_seek_frame(pFormatCtx, video_stream,
-                                         (long) ((double) stream->duration * 0.10), 0);
-                if (seek_ret == 0) {
-                    break;
-                }
+        int video_duration_in_seconds = (int) (pFormatCtx->duration / AV_TIME_BASE);
+
+        int thumbnails_to_generate = (IS_VIDEO(pFormatCtx) && stream->codecpar->codec_id != AV_CODEC_ID_GIF && video_duration_in_seconds >= 15)
+                                     // Limit to ~1 thumbnail every 5s
+                                     ? MAX(MIN(ctx->tn_count, video_duration_in_seconds / 5 + 1), 1) + 1
+                                     : 1;
+
+        const double seek_increment = thumbnails_to_generate == 1
+                                      ? 0.10
+                                      : 1.0 / (thumbnails_to_generate + 1);
+
+        int number_of_thumbnails_generated = 0;
+        int save_thumbnail_ret;
+
+        for (int i = 0; i < thumbnails_to_generate; i++) {
+            double seek_ratio = seek_increment * i + seek_increment * 0.9;
+
+            save_thumbnail_ret = decode_frame_and_save_thumbnail(ctx, pFormatCtx, decoder, stream, video_stream, doc,
+                                                                 seek_ratio, i);
+            if (save_thumbnail_ret == SAVE_THUMBNAIL_FAILED) {
+                break;
+            }
+
+            if (save_thumbnail_ret == SAVE_THUMBNAIL_OK) {
+                number_of_thumbnails_generated += 1;
             }
         }
 
-        frame_and_packet_t *frame_and_packet = read_frame(ctx, pFormatCtx, decoder, video_stream, doc);
-        if (frame_and_packet == NULL) {
-            avcodec_free_context(&decoder);
-            avformat_close_input(&pFormatCtx);
-            avformat_free_context(pFormatCtx);
-            return;
+        if (number_of_thumbnails_generated > 0) {
+            APPEND_LONG_META(doc, MetaThumbnail, number_of_thumbnails_generated)
         }
 
-        if (ctx->tesseract_lang != NULL && STREAM_IS_IMAGE) {
-            ocr_image(ctx, doc, decoder, frame_and_packet->frame);
-        }
-
-        // NOTE: OCR'd content takes precedence over exif image description
-        append_video_meta(ctx, pFormatCtx, frame_and_packet->frame, doc, IS_VIDEO(pFormatCtx));
-
-        // Scale frame
-        AVFrame *scaled_frame = scale_frame(decoder, frame_and_packet->frame, ctx->tn_size);
-
-        if (scaled_frame == NULL) {
-            frame_and_packet_free(frame_and_packet);
-            avcodec_free_context(&decoder);
-            avformat_close_input(&pFormatCtx);
-            avformat_free_context(pFormatCtx);
-            return;
-        }
-
-        if (scaled_frame == STORE_AS_IS) {
-            APPEND_TN_META(doc, frame_and_packet->frame->width, frame_and_packet->frame->height)
-            ctx->store((char *) doc->path_md5, sizeof(doc->path_md5), (char *) frame_and_packet->packet->data,
-                       frame_and_packet->packet->size);
-        } else {
-            // Encode frame to jpeg
-            AVCodecContext *jpeg_encoder = alloc_jpeg_encoder(scaled_frame->width, scaled_frame->height,
-                                                              ctx->tn_qscale);
-            avcodec_send_frame(jpeg_encoder, scaled_frame);
-
-            AVPacket jpeg_packet;
-            av_init_packet(&jpeg_packet);
-            avcodec_receive_packet(jpeg_encoder, &jpeg_packet);
-
-            // Save thumbnail
-            APPEND_TN_META(doc, scaled_frame->width, scaled_frame->height)
-            ctx->store((char *) doc->path_md5, sizeof(doc->path_md5), (char *) jpeg_packet.data, jpeg_packet.size);
-
-            avcodec_free_context(&jpeg_encoder);
-            av_packet_unref(&jpeg_packet);
-            av_free(*scaled_frame->data);
-            av_frame_free(&scaled_frame);
-        }
-
-        frame_and_packet_free(frame_and_packet);
         avcodec_free_context(&decoder);
     }
 
@@ -772,7 +844,7 @@ int store_image_thumbnail(scan_media_ctx_t *ctx, void *buf, size_t buf_len, docu
     }
 
     if (scaled_frame == STORE_AS_IS) {
-        APPEND_TN_META(doc, frame_and_packet->frame->width, frame_and_packet->frame->height)
+        APPEND_LONG_META(doc, MetaThumbnail, 1)
         ctx->store((char *) doc->path_md5, sizeof(doc->path_md5), (char *) frame_and_packet->packet->data,
                    frame_and_packet->packet->size);
     } else {
@@ -786,7 +858,7 @@ int store_image_thumbnail(scan_media_ctx_t *ctx, void *buf, size_t buf_len, docu
         avcodec_receive_packet(jpeg_encoder, &jpeg_packet);
 
         // Save thumbnail
-        APPEND_TN_META(doc, scaled_frame->width, scaled_frame->height)
+        APPEND_LONG_META(doc, MetaThumbnail, 1)
         ctx->store((char *) doc->path_md5, sizeof(doc->path_md5), (char *) jpeg_packet.data, jpeg_packet.size);
 
         av_packet_unref(&jpeg_packet);

third-party/libscan/libscan/media/media.h

@@ -17,6 +17,9 @@ typedef struct {
 
     int tn_size;
     float tn_qscale;
+    /** Number of thumbnails to generate for videos */
+    int tn_count;
+
     long max_media_buffer;
     int read_subtitles;
 

third-party/libscan/libscan/msdoc/msdoc.c

@@ -76,6 +76,7 @@ void parse_msdoc_pdf(scan_msdoc_ctx_t *ctx, document_t *doc, FILE *file, void *b
     scan_ebook_ctx_t ebook_ctx = {
             .content_size = ctx->content_size,
             .tn_size = ctx->tn_size,
+            .enable_tn = TRUE,
             .log = ctx->log,
             .logf = ctx->logf,
             .store = ctx->store,
@@ -137,7 +138,7 @@ void parse_msdoc(scan_msdoc_ctx_t *ctx, vfile_t *f, document_t *doc) {
         return;
     }
 
-    if (ctx->tn_size > 0) {
+    if (ctx->enable_tn) {
         char *buf_pdf = malloc(buf_len);
         memcpy(buf_pdf, buf, buf_len);
         parse_msdoc_pdf(ctx, doc, file, buf_pdf, buf_len);

third-party/libscan/libscan/msdoc/msdoc.h

@@ -5,6 +5,7 @@
 
 typedef struct {
     long content_size;
+    int enable_tn;
     int tn_size;
     log_callback_t log;
     logf_callback_t logf;

third-party/libscan/libscan/ooxml/ooxml.c

@@ -190,7 +190,7 @@ void read_thumbnail(scan_ooxml_ctx_t *ctx, document_t *doc, struct archive *a, s
     char *buf = malloc(entry_size);
     archive_read_data(a, buf, entry_size);
 
-    APPEND_TN_META(doc, 1, 1) // Size unknown
+    APPEND_LONG_META(doc, MetaThumbnail, 1)
     ctx->store((char *) doc->path_md5, sizeof(doc->path_md5), buf, entry_size);
     free(buf);
 }
@@ -238,7 +238,7 @@ void parse_ooxml(scan_ooxml_ctx_t *ctx, vfile_t *f, document_t *doc) {
                 if (read_doc_props(ctx, a, doc) != 0) {
                     break;
                 }
-            } else if (strcmp(path, "docProps/thumbnail.jpeg") == 0) {
+            } else if (ctx->enable_tn && strcmp(path, "docProps/thumbnail.jpeg") == 0) {
                 read_thumbnail(ctx, doc, a, entry);
             }
         }

third-party/libscan/libscan/ooxml/ooxml.h

@@ -5,6 +5,7 @@
 #include "../scan.h"
 
 typedef struct {
+    int enable_tn;
     long content_size;
     log_callback_t log;
     logf_callback_t logf;

third-party/libscan/libscan/raw/raw.c

@@ -69,7 +69,7 @@ int store_thumbnail_rgb24(scan_raw_ctx_t *ctx, libraw_processed_image_t *img, do
     av_init_packet(&jpeg_packet);
     avcodec_receive_packet(jpeg_encoder, &jpeg_packet);
 
-    APPEND_TN_META(doc, scaled_frame->width, scaled_frame->height)
+    APPEND_LONG_META(doc, MetaThumbnail, 1)
     ctx->store((char *) doc->path_md5, sizeof(doc->path_md5), (char *) jpeg_packet.data, jpeg_packet.size);
 
     av_packet_unref(&jpeg_packet);
@@ -157,7 +157,7 @@ void parse_raw(scan_raw_ctx_t *ctx, vfile_t *f, document_t *doc) {
 
     APPEND_STR_META(doc, MetaMediaVideoCodec, "raw")
 
-    if (ctx->tn_size <= 0) {
+    if (!ctx->enable_tn) {
         free(buf);
         libraw_close(libraw_lib);
         return;

third-party/libscan/libscan/raw/raw.h

@@ -8,6 +8,7 @@ typedef struct {
     logf_callback_t logf;
     store_callback_t store;
 
+    int enable_tn;
     int tn_size;
     float tn_qscale;
 } scan_raw_ctx_t;

third-party/libscan/test/main.cpp

@@ -350,9 +350,13 @@ TEST(Comic, ComicIssue160) {
     load_doc_file("libscan-test-files/test_files/ebook/comic-segfault-issue-160.cbr", &f, &doc);
 
     int tn_size_saved = comic_ctx.tn_size;
-    comic_ctx.tn_size = 0;
+    size_t size_before = store_size;
+
+    comic_ctx.enable_tn = FALSE;
     parse_comic(&comic_ctx, &f, &doc);
-    comic_ctx.tn_size = tn_size_saved;
+    comic_ctx.enable_tn = tn_size_saved;
+
+    ASSERT_EQ(store_size, size_before);
 
     cleanup(&doc, &f);
 }
@@ -669,8 +673,6 @@ TEST(Ooxml, Docx2Archive) {
     ASSERT_EQ(get_meta(&LastSubDoc, MetaPages)->long_val, 1);
     ASSERT_EQ(strlen(get_meta(&LastSubDoc, MetaContent)->str_val), 2780);
 
-    fprintf(stderr, "%s\n", get_meta(&LastSubDoc, MetaContent)->str_val);
-
     ooxml_500_ctx.content_size = 500;
 
     cleanup(&doc, &f);
@@ -1111,6 +1113,7 @@ int main(int argc, char **argv) {
     ebook_ctx.tesseract_lang = "eng";
     ebook_ctx.tesseract_path = "./tessdata";
     ebook_ctx.tn_size = 500;
+    ebook_ctx.enable_tn = TRUE;
     ebook_ctx.log = noop_log;
     ebook_ctx.logf = noop_logf;
     ebook_ctx.fast_epub_parse = 0;
@@ -1124,12 +1127,14 @@ int main(int argc, char **argv) {
 
     comic_ctx.tn_qscale = 1.0;
     comic_ctx.tn_size = 500;
+    comic_ctx.enable_tn = TRUE;
     comic_ctx.log = noop_log;
     comic_ctx.logf = noop_logf;
     comic_ctx.store = counter_store;
 
     comic_big_ctx.tn_qscale = 1.0;
     comic_big_ctx.tn_size = 5000;
+    comic_big_ctx.enable_tn = TRUE;
     comic_big_ctx.log = noop_log;
     comic_big_ctx.logf = noop_logf;
     comic_big_ctx.store = counter_store;
@@ -1138,10 +1143,12 @@ int main(int argc, char **argv) {
     media_ctx.logf = noop_logf;
     media_ctx.store = counter_store;
     media_ctx.tn_size = 500;
+    media_ctx.tn_count = 1;
     media_ctx.tn_qscale = 1.0;
     media_ctx.max_media_buffer = (long) 2000 * (long) 1024 * (long) 1024;
 
     ooxml_500_ctx.content_size = 500;
+    ooxml_500_ctx.enable_tn = TRUE;
     ooxml_500_ctx.log = noop_log;
     ooxml_500_ctx.logf = noop_logf;
     ooxml_500_ctx.store = counter_store;
@@ -1154,6 +1161,7 @@ int main(int argc, char **argv) {
     raw_ctx.logf = noop_logf;
     raw_ctx.store = counter_store;
     raw_ctx.tn_size = 500;
+    raw_ctx.enable_tn = TRUE;
     raw_ctx.tn_qscale = 5.0;
 
     msdoc_ctx.log = noop_log;
@@ -1161,12 +1169,14 @@ int main(int argc, char **argv) {
     msdoc_ctx.store = counter_store;
     msdoc_ctx.content_size = 500;
     msdoc_ctx.tn_size = 500;
+    msdoc_ctx.enable_tn = TRUE;
 
     msdoc_text_ctx.log = noop_log;
     msdoc_text_ctx.logf = noop_logf;
     msdoc_text_ctx.store = counter_store;
     msdoc_text_ctx.content_size = 500;
     msdoc_text_ctx.tn_size = 0;
+    msdoc_text_ctx.enable_tn = FALSE;
 
     wpd_ctx.log = noop_log;
     wpd_ctx.logf = noop_logf;