ebook component (with mupdf workaround)

.gitignore

@@ -6,4 +6,5 @@ libscan.so
 *.cbp
 CMakeFiles
 CMakeCache.txt
-scan_test
+scan_test
+third-party/

CMakeLists.txt

@@ -4,6 +4,32 @@ project(scan C)
 set(CMAKE_C_STANDARD 11)
 
 find_package(LibArchive REQUIRED)
+find_package(Threads REQUIRED)
+find_package(Tesseract CONFIG REQUIRED)
+find_package(harfbuzz CONFIG REQUIRED)
+find_package(OpenJPEG CONFIG REQUIRED)
+find_package(JPEG REQUIRED)
+
+
+include(ExternalProject)
+find_program(MAKE_EXE NAMES gmake nmake make)
+ExternalProject_Add(
+        mupdf
+        # TODO: use master branch ?
+        URL https://mupdf.com/downloads/archive/mupdf-1.16.1-source.tar.xz
+
+        UPDATE_COMMAND ""
+        PATCH_COMMAND ""
+        TEST_COMMAND ""
+        CONFIGURE_COMMAND ""
+        INSTALL_COMMAND ""
+
+        PREFIX "third-party/ext_mupdf"
+        BINARY_DIR "third-party/ext_mupdf/src/mupdf"
+
+        BUILD_COMMAND CFLAGS=-fPIC HAVE_CURL=no HAVE_GLUT=no ${MAKE_EXE} -j 4 --silent
+            && ar d build/release/libmupdf-third.a jutils.o jdinput.o jdmarker.o jdmaster.o
+)
 
 add_library(
         scan
@@ -13,6 +39,7 @@ add_library(
 
         libscan/text/text.c libscan/text/text.h
         libscan/arc/arc.c libscan/arc/arc.h
+        libscan/ebook/ebook.c libscan/ebook/ebook.h
 
         third-party/utf8.h
 )
@@ -24,10 +51,42 @@ target_compile_options(
         -g
 )
 
+add_dependencies(
+        scan
+        mupdf
+)
+
+SET(CMAKE_C_LINK_EXECUTABLE "g++ <FLAGS> <CMAKE_C_LINK_FLAGS> <LINK_FLAGS> <OBJECTS> -o <TARGET> <LINK_LIBRARIES>")
+
 target_link_libraries(
         scan
+
         -static
+        -static-libgcc
+        -static-libstdc++
+
+        -Wl,--whole-archive
+        m
+        -Wl,--no-whole-archive
+
+        "${CMAKE_SOURCE_DIR}/third-party/ext_mupdf/src/mupdf/build/release/libmupdf.a"
+        "${CMAKE_SOURCE_DIR}/third-party/ext_mupdf/src/mupdf/build/release/libmupdf-third.a"
+
+        ${JPEG_LIBRARIES}
         ${LibArchive_LIBRARIES}
+        ${Tesseract_LIBRARIES}
+
+        ${CMAKE_THREAD_LIBS_INIT}
+
+        # TODO: Looks like I don't need to explicitly link to libuuid?
+)
+
+target_include_directories(
+        scan
+        BEFORE
+        PRIVATE
+        "${CMAKE_SOURCE_DIR}/third-party/ext_mupdf/src/mupdf/include/"
+        ${JPEG_INCLUDE_DIR}
 )
 
 
@@ -37,7 +96,6 @@ add_executable(
         test/main.c
 )
 
-
 target_link_libraries(
         scan_test
         scan

README.md

@@ -1,4 +1,8 @@
 
 ```bash
-vcpkg install libarchive
+vcpkg install libarchive pthread tesseract
+
+rm -rf CMakeFiles/ CMakeCache.txt
+cmake -DCMAKE_TOOLCHAIN_FILE=/usr/share/vcpkg/scripts/buildsystems/vcpkg.cmake .
+make -j 4
 ```

build.sh

@@ -0,0 +1,8 @@
+#!/usr/bin/env bash
+
+export CC=gcc
+export CXX=g++
+
+rm -rf CMakeFiles CMakeCache.txt
+cmake -DCMAKE_TOOLCHAIN_FILE=/usr/share/vcpkg/scripts/buildsystems/vcpkg.cmake . || exit
+make -j 4

libscan/ebook/ebook.c

@@ -0,0 +1,334 @@
+#include "ebook.h"
+#include "../util.h"
+#include <mupdf/fitz.h>
+#include <pthread.h>
+#include <tesseract/capi.h>
+
+#define MIN_OCR_SIZE 350
+#define MIN_OCR_LEN 10
+
+/* fill_image callback doesn't let us pass opaque pointers unless I create my own device */
+__thread text_buffer_t thread_buffer;
+__thread scan_ebook_ctx_t thread_ctx;
+
+
+int render_cover(scan_ebook_ctx_t *ctx, fz_context *fzctx, document_t *doc, fz_document *fzdoc) {
+
+    int err = 0;
+    fz_page *cover = NULL;
+
+    fz_var(cover);
+    fz_var(err);
+    fz_try(fzctx)
+        cover = fz_load_page(fzctx, fzdoc, 0);
+    fz_catch(fzctx)
+        err = 1;
+
+    if (err != 0) {
+        fz_drop_page(fzctx, cover);
+//        LOG_WARNINGF(doc->filepath, "fz_load_page() returned error code [%d] %s", err, ctx->error.message)
+        return FALSE;
+    }
+
+    fz_rect bounds = fz_bound_page(fzctx, cover);
+
+    float scale;
+    float w = (float) bounds.x1 - bounds.x0;
+    float h = (float) bounds.y1 - bounds.y0;
+    if (w > h) {
+        scale = (float) ctx->tn_size / w;
+    } else {
+        scale = (float) ctx->tn_size / h;
+    }
+    fz_matrix m = fz_scale(scale, scale);
+
+    bounds = fz_transform_rect(bounds, m);
+    fz_irect bbox = fz_round_rect(bounds);
+    fz_pixmap *pixmap = fz_new_pixmap_with_bbox(fzctx, fzctx->colorspace->rgb, bbox, NULL, 0);
+
+    fz_clear_pixmap_with_value(fzctx, pixmap, 0xFF);
+    fz_device *dev = fz_new_draw_device(fzctx, m, pixmap);
+
+    fz_var(err);
+    fz_try(fzctx)
+    {
+        pthread_mutex_lock(&ctx->mupdf_mutex);
+        fz_run_page(fzctx, cover, dev, fz_identity, NULL);
+    }
+    fz_always(fzctx)
+    {
+        fz_close_device(fzctx, dev);
+        fz_drop_device(fzctx, dev);
+        pthread_mutex_unlock(&ctx->mupdf_mutex);
+    }
+    fz_catch(fzctx)
+        err = fzctx->error.errcode;
+
+    if (err != 0) {
+//        LOG_WARNINGF(doc->filepath, "fz_run_page() returned error code [%d] %s", err, ctx->error.message)
+        fz_drop_page(fzctx, cover);
+        fz_drop_pixmap(fzctx, pixmap);
+        return FALSE;
+    }
+
+    fz_buffer *fzbuf = NULL;
+    fz_var(fzbuf);
+    fz_var(err);
+
+    fz_try(fzctx)
+        fzbuf = fz_new_buffer_from_pixmap_as_png(fzctx, pixmap, fz_default_color_params);
+    fz_catch(fzctx)
+        err = fzctx->error.errcode;
+
+    if (err == 0) {
+        unsigned char *tn_buf;
+        size_t tn_len = fz_buffer_storage(fzctx, fzbuf, &tn_buf);
+//        store_write(ScanCtx.index.store, (char *) doc->uuid, sizeof(doc->uuid), (char *) tn_buf, tn_len);
+    }
+
+    fz_drop_buffer(fzctx, fzbuf);
+    fz_drop_pixmap(fzctx, pixmap);
+    fz_drop_page(fzctx, cover);
+
+    if (err != 0) {
+//        LOG_WARNINGF(doc->filepath, "fz_new_buffer_from_pixmap_as_png() returned error code [%d] %s", err,
+//                     ctx->error.message)
+        return FALSE;
+    }
+
+    return TRUE;
+}
+
+void fz_err_callback(void *user, UNUSED(const char *message)) {
+//    if (LogCtx.verbose) {
+//        document_t *doc = (document_t *) user;
+//        LOG_WARNINGF(doc->filepath, "FZ: %s", message)
+//    }
+}
+
+static void init_fzctx(fz_context *fzctx, document_t *doc) {
+    fz_disable_icc(fzctx);
+    fz_register_document_handlers(fzctx);
+
+    fzctx->warn.print_user = doc;
+    fzctx->warn.print = fz_err_callback;
+    fzctx->error.print_user = doc;
+    fzctx->error.print = fz_err_callback;
+}
+
+static int read_stext_block(fz_stext_block *block, text_buffer_t *tex) {
+    if (block->type != FZ_STEXT_BLOCK_TEXT) {
+        return 0;
+    }
+
+    fz_stext_line *line = block->u.t.first_line;
+    while (line != NULL) {
+        fz_stext_char *c = line->first_char;
+        while (c != NULL) {
+            if (text_buffer_append_char(tex, c->c) == TEXT_BUF_FULL) {
+                return TEXT_BUF_FULL;
+            }
+            c = c->next;
+        }
+        line = line->next;
+    }
+    return 0;
+}
+
+#define IS_VALID_BPP(d) (d==1 || d==2 || d==4 || d==8 || d==16 || d==24 || d==32)
+
+void fill_image(fz_context *fzctx, UNUSED(fz_device *dev),
+                fz_image *img, UNUSED(fz_matrix ctm), UNUSED(float alpha),
+                UNUSED(fz_color_params color_params)) {
+
+    int l2factor = 0;
+
+    if (img->w > MIN_OCR_SIZE && img->h > MIN_OCR_SIZE && IS_VALID_BPP(img->n)) {
+
+        fz_pixmap *pix = img->get_pixmap(fzctx, img, NULL, img->w, img->h, &l2factor);
+
+        if (pix->h > MIN_OCR_SIZE && img->h > MIN_OCR_SIZE && img->xres != 0) {
+            TessBaseAPI *api = TessBaseAPICreate();
+            TessBaseAPIInit3(api, thread_ctx.tesseract_path, thread_ctx.tesseract_lang);
+
+            TessBaseAPISetImage(api, pix->samples, pix->w, pix->h, pix->n, pix->stride);
+            TessBaseAPISetSourceResolution(api, pix->xres);
+
+            char *text = TessBaseAPIGetUTF8Text(api);
+            size_t len = strlen(text);
+            if (len >= MIN_OCR_LEN) {
+                text_buffer_append_string(&thread_buffer, text, len - 1);
+//                LOG_DEBUGF(
+//                        "pdf.c",
+//                        "(OCR) %dx%d got %dB from tesseract (%s), buffer:%dB",
+//                        pix->w, pix->h, len, ScanCtx.tesseract_lang, thread_buffer.dyn_buffer.cur
+//                )
+            }
+
+            TessBaseAPIEnd(api);
+            TessBaseAPIDelete(api);
+        }
+        fz_drop_pixmap(fzctx, pix);
+    }
+}
+
+void parse_ebook(scan_ebook_ctx_t *ctx, vfile_t *f, const char* mime_str,  document_t *doc) {
+
+    size_t buf_len;
+    void * buf = read_all(f, &buf_len);
+
+    static int mu_is_initialized = 0;
+    if (!mu_is_initialized) {
+        pthread_mutex_init(&ctx->mupdf_mutex, NULL);
+        mu_is_initialized = 1;
+    }
+    fz_context *fzctx = fz_new_context(NULL, NULL, FZ_STORE_UNLIMITED);
+
+    init_fzctx(fzctx, doc);
+
+    int err = 0;
+
+    fz_document *fzdoc = NULL;
+    fz_stream *stream = NULL;
+    fz_var(fzdoc);
+    fz_var(stream);
+    fz_var(err);
+
+    fz_try(fzctx)
+    {
+        stream = fz_open_memory(fzctx, buf, buf_len);
+        fzdoc = fz_open_document_with_stream(fzctx, mime_str, stream);
+    }
+    fz_catch(fzctx)
+        err = fzctx->error.errcode;
+
+    if (err != 0) {
+        fz_drop_stream(fzctx, stream);
+        fz_drop_document(fzctx, fzdoc);
+        fz_drop_context(fzctx);
+        return;
+    }
+
+    char title[4096] = {'\0',};
+    fz_try(fzctx)
+        fz_lookup_metadata(fzctx, fzdoc, FZ_META_INFO_TITLE, title, sizeof(title));
+    fz_catch(fzctx)
+        ;
+
+    if (strlen(title) > 0) {
+        meta_line_t *meta_content = malloc(sizeof(meta_line_t) + strlen(title));
+        meta_content->key = MetaTitle;
+        strcpy(meta_content->str_val, title);
+        APPEND_META(doc, meta_content)
+    }
+
+    int page_count = -1;
+    fz_var(err);
+    fz_try(fzctx)
+        page_count = fz_count_pages(fzctx, fzdoc);
+    fz_catch(fzctx)
+        err = fzctx->error.errcode;
+
+    if (err) {
+//        LOG_WARNINGF(doc->filepath, "fz_count_pages() returned error code [%d] %s", err, ctx->error.message)
+        fz_drop_stream(fzctx, stream);
+        fz_drop_document(fzctx, fzdoc);
+        fz_drop_context(fzctx);
+        return;
+    }
+
+    if (ctx->tn_size > 0) {
+        err = render_cover(ctx, fzctx, doc, fzdoc);
+    }
+
+    if (err == TRUE) {
+        fz_drop_stream(fzctx, stream);
+        fz_drop_document(fzctx, fzdoc);
+        fz_drop_context(fzctx);
+        return;
+    }
+
+    if (ctx->content_size > 0) {
+        fz_stext_options opts = {0};
+        thread_buffer = text_buffer_create(ctx->content_size);
+
+        for (int current_page = 0; current_page < page_count; current_page++) {
+            fz_page *page = NULL;
+            fz_var(err);
+            fz_try(fzctx)
+                page = fz_load_page(fzctx, fzdoc, current_page);
+            fz_catch(fzctx)
+                err = fzctx->error.errcode;
+            if (err != 0) {
+//                LOG_WARNINGF(doc->filepath, "fz_load_page() returned error code [%d] %s", err, ctx->error.message)
+                text_buffer_destroy(&thread_buffer);
+                fz_drop_page(fzctx, page);
+                fz_drop_stream(fzctx, stream);
+                fz_drop_document(fzctx, fzdoc);
+                fz_drop_context(fzctx);
+                return;
+            }
+
+            fz_stext_page *stext = fz_new_stext_page(fzctx, fz_bound_page(fzctx, page));
+            fz_device *dev = fz_new_stext_device(fzctx, stext, &opts);
+            dev->stroke_path = NULL;
+            dev->stroke_text = NULL;
+            dev->clip_text = NULL;
+            dev->clip_stroke_path = NULL;
+            dev->clip_stroke_text = NULL;
+
+            if (ctx->tesseract_lang!= NULL) {
+                dev->fill_image = fill_image;
+            }
+
+            fz_var(err);
+            fz_try(fzctx)
+                fz_run_page(fzctx, page, dev, fz_identity, NULL);
+            fz_always(fzctx)
+            {
+                fz_close_device(fzctx, dev);
+                fz_drop_device(fzctx, dev);
+            }
+            fz_catch(fzctx)
+                err = fzctx->error.errcode;
+
+            if (err != 0) {
+//                LOG_WARNINGF(doc->filepath, "fz_run_page() returned error code [%d] %s", err, ctx->error.message)
+                text_buffer_destroy(&thread_buffer);
+                fz_drop_page(fzctx, page);
+                fz_drop_stext_page(fzctx, stext);
+                fz_drop_stream(fzctx, stream);
+                fz_drop_document(fzctx, fzdoc);
+                fz_drop_context(fzctx);
+                return;
+            }
+
+            fz_stext_block *block = stext->first_block;
+            while (block != NULL) {
+                int ret = read_stext_block(block, &thread_buffer);
+                if (ret == TEXT_BUF_FULL) {
+                    break;
+                }
+                block = block->next;
+            }
+            fz_drop_stext_page(fzctx, stext);
+            fz_drop_page(fzctx, page);
+
+            if (thread_buffer.dyn_buffer.cur >= thread_buffer.dyn_buffer.size) {
+                break;
+            }
+        }
+        text_buffer_terminate_string(&thread_buffer);
+
+        meta_line_t *meta_content = malloc(sizeof(meta_line_t) + thread_buffer.dyn_buffer.cur);
+        meta_content->key = MetaContent;
+        memcpy(meta_content->str_val, thread_buffer.dyn_buffer.buf, thread_buffer.dyn_buffer.cur);
+        APPEND_META(doc, meta_content)
+
+        text_buffer_destroy(&thread_buffer);
+    }
+
+    fz_drop_stream(fzctx, stream);
+    fz_drop_document(fzctx, fzdoc);
+    fz_drop_context(fzctx);
+}

libscan/ebook/ebook.h

@@ -0,0 +1,16 @@
+#ifndef SCAN_PDF_H
+#define SCAN_PDF_H
+
+#include "../scan.h"
+
+typedef struct {
+    long content_size;
+    int tn_size;
+    const char *tesseract_lang;
+    const char *tesseract_path;
+    pthread_mutex_t mupdf_mutex;
+} scan_ebook_ctx_t;
+
+void parse_ebook(scan_ebook_ctx_t *ctx, vfile_t *f, const char* mime_str,  document_t *doc);
+
+#endif

libscan/macros.h

@@ -1,5 +1,6 @@
 #ifndef	FALSE
 #define	FALSE	(0)
+#define BOOL int
 #endif
 
 #ifndef	TRUE

libscan/scan.h

@@ -10,6 +10,8 @@
 #define META_STR_MASK 0x40
 #define META_LONG_MASK 0x20
 
+#define UNUSED(x) __attribute__((__unused__))  x
+
 #define META_STR(id) ((unsigned) id) | ((unsigned) META_STR_MASK)
 #define META_INT(id) ((unsigned) id) | ((unsigned) META_INT_MASK)
 #define META_LONG(id) ((unsigned) id) | ((unsigned) META_LONG_MASK)
@@ -89,6 +91,7 @@ typedef struct vfile {
 
     int is_fs_file;
     char *filepath;
+    struct stat info;
 
     read_func_t read;
     close_func_t close;

libscan/util.h

@@ -1,9 +1,9 @@
 #ifndef SIST2_UTIL_H
 #define SIST2_UTIL_H
 
-#include <stdio.h>
+#include "stdio.h"
-#include <stdlib.h>
+#include "stdlib.h"
-#include <string.h>
+#include "string.h"
 #include "../third-party/utf8.h/utf8.h"
 #include "macros.h"
 
@@ -26,7 +26,6 @@ typedef struct text_buffer {
     dyn_buffer_t dyn_buffer;
 } text_buffer_t;
 
-__always_inline
 static int utf8_validchr2(const char *s) {
     if (0x00 == (0x80 & *s)) {
         return TRUE;
@@ -75,7 +74,6 @@ static int utf8_validchr2(const char *s) {
 }
 
 
-__always_inline
 static dyn_buffer_t dyn_buffer_create() {
     dyn_buffer_t buf;
 
@@ -86,7 +84,6 @@ static dyn_buffer_t dyn_buffer_create() {
     return buf;
 }
 
-__always_inline
 static void grow_buffer(dyn_buffer_t *buf, size_t size) {
     if (buf->cur + size > buf->size) {
         do {
@@ -97,7 +94,6 @@ static void grow_buffer(dyn_buffer_t *buf, size_t size) {
     }
 }
 
-__always_inline
 static void grow_buffer_small(dyn_buffer_t *buf) {
     if (buf->cur + sizeof(long) > buf->size) {
         buf->size *= 2;
@@ -105,7 +101,6 @@ static void grow_buffer_small(dyn_buffer_t *buf) {
     }
 }
 
-__always_inline
 static void dyn_buffer_write(dyn_buffer_t *buf, const void *data, size_t size) {
     grow_buffer(buf, size);
 
@@ -113,7 +108,6 @@ static void dyn_buffer_write(dyn_buffer_t *buf, const void *data, size_t size) {
     buf->cur += size;
 }
 
-__always_inline
 static void dyn_buffer_write_char(dyn_buffer_t *buf, char c) {
     grow_buffer_small(buf);
 
@@ -121,18 +115,15 @@ static void dyn_buffer_write_char(dyn_buffer_t *buf, char c) {
     buf->cur += sizeof(c);
 }
 
-__always_inline
 static void dyn_buffer_write_str(dyn_buffer_t *buf, char *str) {
     dyn_buffer_write(buf, str, strlen(str));
     dyn_buffer_write_char(buf, '\0');
 }
 
-__always_inline
 static void dyn_buffer_append_string(dyn_buffer_t *buf, char *str) {
     dyn_buffer_write(buf, str, strlen(str));
 }
 
-__always_inline
 static void dyn_buffer_write_int(dyn_buffer_t *buf, int d) {
     grow_buffer_small(buf);
 
@@ -140,7 +131,6 @@ static void dyn_buffer_write_int(dyn_buffer_t *buf, int d) {
     buf->cur += sizeof(int);
 }
 
-__always_inline
 static void dyn_buffer_write_short(dyn_buffer_t *buf, short s) {
     grow_buffer_small(buf);
 
@@ -148,7 +138,6 @@ static void dyn_buffer_write_short(dyn_buffer_t *buf, short s) {
     buf->cur += sizeof(short);
 }
 
-__always_inline
 static void dyn_buffer_write_long(dyn_buffer_t *buf, unsigned long l) {
     grow_buffer_small(buf);
 
@@ -156,17 +145,14 @@ static void dyn_buffer_write_long(dyn_buffer_t *buf, unsigned long l) {
     buf->cur += sizeof(unsigned long);
 }
 
-__always_inline
 static void dyn_buffer_destroy(dyn_buffer_t *buf) {
     free(buf->buf);
 }
 
-__always_inline
 static void text_buffer_destroy(text_buffer_t *buf) {
     dyn_buffer_destroy(&buf->dyn_buffer);
 }
 
-__always_inline
 static text_buffer_t text_buffer_create(long max_size) {
     text_buffer_t text_buf;
 
@@ -177,7 +163,6 @@ static text_buffer_t text_buffer_create(long max_size) {
     return text_buf;
 }
 
-__always_inline
 static int text_buffer_append_char(text_buffer_t *buf, int c) {
 
     if (SHOULD_IGNORE_CHAR(c) || c == ' ') {
@@ -218,7 +203,6 @@ static int text_buffer_append_char(text_buffer_t *buf, int c) {
 }
 
 
-__always_inline
 static void text_buffer_terminate_string(text_buffer_t *buf) {
     if (buf->dyn_buffer.cur > 0 && *(buf->dyn_buffer.buf + buf->dyn_buffer.cur - 1) == ' ') {
         *(buf->dyn_buffer.buf + buf->dyn_buffer.cur - 1) = '\0';
@@ -233,7 +217,6 @@ static void text_buffer_terminate_string(text_buffer_t *buf) {
     (0xe0 == (0xf0 & *ptr) && ptr - str > len - 3) || \
     (0xf0 == (0xf8 & *ptr) && ptr - str > len - 4))
 
-__always_inline
 static int text_buffer_append_string(text_buffer_t *buf, const char *str, size_t len) {
 
     const char *ptr = str;
@@ -275,9 +258,17 @@ static int text_buffer_append_string(text_buffer_t *buf, const char *str, size_t
     return 0;
 }
 
-__always_inline
 static int text_buffer_append_string0(text_buffer_t *buf, char *str) {
     return text_buffer_append_string(buf, str, strlen(str));
 }
 
+static void* read_all(vfile_t *f, size_t *size) {
+    void* buf = malloc(f->info.st_size);
+    *size = f->read(f, buf, f->info.st_size);
+
+    //TODO: log
+
+    return buf;
+}
+
 #endif

test/main.c

@@ -1,10 +1,11 @@
 #include "../libscan/text/text.h"
 #include <fcntl.h>
 #include "../libscan/arc/arc.h"
+#include "../libscan/ebook/ebook.h"
 
 int main() {
 
-    scan_text_ctx_t ctx;
+    scan_ebook_ctx_t ctx;
 
     ctx.content_size = 100;
     vfile_t file;
@@ -18,5 +19,5 @@ int main() {
     doc.meta_tail = NULL;
 
     doc.size = 200;
-    parse_text(&ctx, &file, &doc);
+    parse_ebook(&ctx,  &file,"application/pdf", &doc);
 }