FB3 format basic support (for QT only)

[pkb]

Open issues

• Advanced FB3 features such as SVG images, float property, extended description etc are not supported
• fb2.css was used for fb3 as is, probably some tweaks will be required for FB3
• Not much tested, https://github.com/gribuser/FB3/blob/master/Examples/Hardcore%20file%20structure.fb3 can't be opened.
• images having names like '%D0%BA%D0%BB%D0%B5%D1%82%D0%BA%D0%B0.jpg' in Anathomy tutorial example.fb3 also can't be opened (getObjectImageRefName() calls DecodeHTMLUrlString() which converts these %* chars, as a result the file can't be found in a zip file, not sure how to handle this caseproperly

[Frenzie]

Pinging @poire-z

@pkb Btw, we use nanosvg for some basic SVG support, but unfortunately I notice the library's added a notice that it's no longer actively maintained.

[pkb]

@Frenzie thaks, I saw it. I assume when #91 issue will be closed we will be using it too :) . Don't know how good is nanosvg, but it is cool it is header only library it is a pity it is no longer maintained,.

[poire-z]

(@Frenzie : looks easily mergable, but let's wait a bit for followup stuff if any - as there's no real demand for fb3 on our side.)

@pkb : regarding your 4th issue:

images having names like '%D0%BA%D0%BB%D0%B5%D1%82%D0%BA%D0%B0.jpg' can't be opened

We may have fixed that in koreader/crengine#143 (small commit koreader/crengine@3a74635 ). Mostly tested with links, but may work with images src.

[pkb]

@poire-z, thanks with this patch I'm able to see proper file name in the error message in the log. The image is in a zip file so it seems I shouldn't call DecodeHTMLUrlString() to open it. I just don't know conditions when to call and when not, so i'm going to leave it as is. Don't think it is a big issue.

[virxkane]

@pkb, just retry without DecodeHTMLUrlString() if file can't be opened. Like this:

diff --git a/crengine/include/lvtinydom.h b/crengine/include/lvtinydom.h
index 30549a092..e5bd05813 100644
--- a/crengine/include/lvtinydom.h
+++ b/crengine/include/lvtinydom.h
@@ -851,7 +851,7 @@ public:
     /// returns object image source
     LVImageSourceRef getObjectImageSource();
     /// returns object image ref name
-    lString16 getObjectImageRefName();
+    lString16 getObjectImageRefName(bool percentDecode = true);
     /// returns object image stream
     LVStreamRef getObjectImageStream();
     /// formats final block
diff --git a/crengine/src/lvtinydom.cpp b/crengine/src/lvtinydom.cpp
index 1f7daa7c7..d96c2c113 100644
--- a/crengine/src/lvtinydom.cpp
+++ b/crengine/src/lvtinydom.cpp
@@ -11035,7 +11035,7 @@ public:
 };
 
 /// returns object image ref name
-lString16 ldomNode::getObjectImageRefName()
+lString16 ldomNode::getObjectImageRefName(bool percentDecode)
 {
     if (!isElement())
         return lString16::empty_str;
@@ -11072,7 +11072,8 @@ lString16 ldomNode::getObjectImageRefName()
     }
     if ( refName.length()<2 )
         return lString16::empty_str;
-    refName = DecodeHTMLUrlString(refName);
+    if (percentDecode)
+        refName = DecodeHTMLUrlString(refName);
     return refName;
 }
 
@@ -11090,11 +11091,18 @@ LVStreamRef ldomNode::getObjectImageStream()
 /// returns object image source
 LVImageSourceRef ldomNode::getObjectImageSource()
 {
-    lString16 refName = getObjectImageRefName();
+    lString16 refName = getObjectImageRefName(true);
     LVImageSourceRef ref;
     if ( refName.empty() )
         return ref;
     ref = getDocument()->getObjectImageSource( refName );
+    if (ref.isNull()) {
+        // try again without percent decoding (for fb3)
+        refName = getObjectImageRefName(false);
+        if ( refName.empty() )
+            return ref;
+        ref = getDocument()->getObjectImageSource( refName );
+    }
     if ( !ref.isNull() ) {
         int dx = ref->GetWidth();
         int dy = ref->GetHeight();

[virxkane]

Why tag "spacing" create new paragraph? It's right?

[pkb]

@virxkane thanks for testing, no it is not right, i will make it inline in fb3.css and will apply your patch for DecodeHTMLUrlString() today.

[virxkane]

Tag "spacing": still new paragraphs created.
Document's language not loaded from description file, may be this patch will be usefull?

diff --git a/crengine/src/fb3fmt.cpp b/crengine/src/fb3fmt.cpp
index 9d2dcd9d5..e0e3bab8c 100644
--- a/crengine/src/fb3fmt.cpp
+++ b/crengine/src/fb3fmt.cpp
@@ -4,6 +4,7 @@
 
 const lChar16 * const fb3_BodyContentType = L"application/fb3-body+xml";
 const lChar16 * const fb3_PropertiesContentType = L"application/vnd.openxmlformats-package.core-properties+xml";
+const lChar16 * const fb3_DescriptionContentType = L"application/fb3-description+xml";
 const lChar16 * const fb3_CoverRelationship = L"http://schemas.openxmlformats.org/package/2006/relationships/metadata/thumbnail";
 const lChar16 * const fb3_ImageRelationship = L"http://www.fictionbook.org/FictionBook3/relationships/image";
 
@@ -128,10 +129,24 @@ bool ImportFb3Document( LVStreamRef stream, ldomDocument * doc, LVDocViewCallbac
         return false;
     }
 
+    LVStreamRef descStream = arc->OpenStream(context.getContentPath(fb3_DescriptionContentType).c_str(), LVOM_READ );
+    if ( descStream.isNull() ) {
+        CRLog::error("Couldn't read description");
+        return false;
+    }
+
+    ldomDocument * descDoc = LVParseXMLStream( descStream );
+    if ( !descDoc ) {
+        CRLog::error("Couldn't parse description doc");
+        return false;
+    }
+
     lString16 author = propertiesDoc->textFromXPath( cs16("coreProperties/creator") );
     lString16 title = doc->textFromXPath( cs16("coreProperties/title") );
+    lString16 language = descDoc->textFromXPath( cs16("fb3-description/lang") );
     doc_props->setString(DOC_PROP_TITLE, title);
     doc_props->setString(DOC_PROP_AUTHORS, author );
+    doc_props->setString(DOC_PROP_LANGUAGE, language);
     CRLog::info("Author: %s Title: %s", author.c_str(), title.c_str());
     delete propertiesDoc;

[pkb]

@virxkane thanks, now spacing should be inline, but it looks like we have a glitch with letter-spacing

[virxkane]

@pkb Can you fix this glitches?

[pkb]

Don't know. I will take a look, if it will not require much efforts I will :).

[virxkane]

@pkb looks good.
Other glitch about footnotes (file nightmare_example.fb3):


Why are there so many footnotes?

[virxkane]

Lists: Invalid item numbers: starting at 2, not 1.

The third item has been moved to the next page, possibly due to a nested tag "title".

P.S. In the file "nightmare_example.fb3" the tags "ol" and "ul" must be replaced, but this is creator's mistake.