Zorba

Code review comment for lp:~zorba-coders/zorba/fread-pdf-trunk

fread-pdf-trunk
Merge into trunk

Revision history for this message

Matthias Brantner (matthias-brantner) wrote on 2012-09-21:

The module works pretty decent. I was able to extract text or generate images for several pdfs without any problems.
There are some minor things that should be discussed and/or fixed:

- the error seems to be too general, essentially it always raises JAVA-EXCEPTION no matter what goes wrong (e.g. it the given input is not a valid pdf)

- the java stack trace seems to be sent to standard error

- Renders the each page of the PDF document as an image. => Renders each page of the PDF document as an image.

- the names of the private functions should also adhere to the code conventions renderToImages => render-to-images

- make xqdoc failes because the comments seem to contain invalid xml
</home/mbrantner/zorba/build/URI_PATH/com/zorba-xquery/www/modules/project_xqdoc.xq>:142,9: user-defined error [err:UE004]: Error processing module zerr:ZXQD0002 - " This module provides funtionality to read the text from PDF documents and
to render PDF documents to images.
<a href="http://pdfbox.apache.org">Apache PDFBox</a> library is used to
implement these functions.
 
 
Note: Since this module has a Java library dependency a JVM required
to be installed on the system. For Windows: jvm.dll is required on the system
path ( usually located in "C:\Program Files\Java\jre6\bin\client".
Note: For Debian based Linux distributions install PdfBox and FontBox
packages: sudo apt-get install libpdfbox-java libfontbox-java
": can not parse as XML for xqdoc: loader parsing error: Opening and ending tag mismatch: b line 0 and root
; raised at /home/mbrantner/zorba/sandbox/src/runtime/errors_and_diagnostics/errors_and_diagnostics_impl.cpp:81

- would it make sense to return one string per page in the pdf instead of one big string?

- remove commented out code in read-pdf.cpp

- valgrind shows tons of invalid writes. Why? Are they critical? Is there anything we can do?

- would it make sense to return the images in a streaming fashion (i.e. don't create all base64's in a vector)?

- encoding each image shouldn't be necessary and will probably we wasted effort because the images might be written to a file in their binary form

The module works pretty decent. I was able to extract text or generate images for several pdfs without any problems.
There are some minor things that should be discussed and/or fixed:

- the error seems to be too general, essentially it always raises JAVA-EXCEPTION no matter what goes wrong (e.g. it the given input is not a valid pdf)

- the java stack trace seems to be sent to standard error

- Renders the each page of the PDF document as an image. => Renders each page of the PDF document as an image.

- the names of the private functions should also adhere to the code conventions renderToImages => render-to-images

- make xqdoc failes because the comments seem to contain invalid xml
</home/mbrantner/zorba/build/URI_PATH/com/zorba-xquery/www/modules/project_xqdoc.xq>:142,9: user-defined error [err:UE004]: Error processing module zerr:ZXQD0002 - " This module provides funtionality to read the text from PDF documents and
 to render PDF documents to images.
 <a href="http://pdfbox.apache.org">Apache PDFBox</a> library is used to
 implement these functions.
 
 
 Note: Since this module has a Java library dependency a JVM required
 to be installed on the system. For Windows: jvm.dll is required on the system
 path ( usually located in "C:\Program Files\Java\jre6\bin\client".
 Note: For Debian based Linux distributions install PdfBox and FontBox
 packages: sudo apt-get install libpdfbox-java libfontbox-java
": can not parse as XML for xqdoc: loader parsing error: Opening and ending tag mismatch: b line 0 and root
; raised at /home/mbrantner/zorba/sandbox/src/runtime/errors_and_diagnostics/errors_and_diagnostics_impl.cpp:81

- would it make sense to return one string per page in the pdf instead of one big string?

- remove commented out code in read-pdf.cpp

- valgrind shows tons of invalid writes. Why? Are they critical? Is there anything we can do?

- would it make sense to return the images in a streaming fashion (i.e. don't create all base64's in a vector)?

- encoding each image shouldn't be necessary and will probably we wasted effort because the images might be written to a file in their binary form

review: Needs Fixing

« Back to merge proposal