We get a large amount of data from our clients in pdf files in varying formats [layout-wise], these files are typically report output, and are typically properly annotated [they don’t usually need OCR], but not formatted well enough that simply copying several hundred pages of text out of acrobat is not going to work.
The best approach I’ve found so far is to write a script to parse the nearly-valid xml output (the comments are invalid and many characters are escaped in varying ways, é becomes [[[e9]]]é, $ becomes \$, % becomes \%…) of the command-line pdftoipe utility (to convert pdf files for a program called ipe), which gives me text elements with their positions on each page [see sample below], which works well enough for reports where the same values are on the same place on every page I care about, but would require extra scripting effort for importing matrix [cross-tab] pdf files. pdftoipe is not at all intended for this, and at best can be compiled manually using cygwin for windows.
Are there libraries that make this easy from some scripting language I can tolerate? A graphical tool would be awesome too. And a pony.
pdftoipe output of this sample looks like this:
<ipe creator='pdftoipe 2006/10/09'><info media='0 0 612 792'/> <-- Page: 1 1 --> <page gridsize='8'> <path fill='1 1 1' fillrule='wind'> 64.8 144 m 486 144 l 486 727.2 l 64.8 727.2 l 64.8 144 l h </path> <path fill='1 1 1' fillrule='wind'> 64.8 144 m 486 144 l 486 727.2 l 64.8 727.2 l 64.8 144 l h </path> <path fill='1 1 1' fillrule='wind'> 64.8 144 m 486 144 l 486 727.2 l 64.8 727.2 l 64.8 144 l h </path> <text stroke='1 0 0' pos='0 0' size='18' transformable='yes' matrix='1 0 0 1 181.8 707.88'>This is a sample PDF fil</text> <text stroke='1 0 0' pos='0 0' size='18' transformable='yes' matrix='1 0 0 1 356.28 707.88'>e.</text> <text stroke='1 0 0' pos='0 0' size='18' transformable='yes' matrix='1 0 0 1 368.76 707.88'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 692.4'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 677.88'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 663.36'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 648.84'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 634.32'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 619.8'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 605.28'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 590.76'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 576.24'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 561.72'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 547.2'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 532.68'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 518.16'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 503.64'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 489.12'> </text> <text stroke='0 0 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 67.32 474.6'> </text> <text stroke='0 0 1' pos='0 0' size='16.2' transformable='yes' matrix='1 0 0 1 67.32 456.24'>If you can read this</text> <text stroke='0 0 1' pos='0 0' size='16.2' transformable='yes' matrix='1 0 0 1 214.92 456.24'>,</text> <text stroke='0 0 1' pos='0 0' size='16.2' transformable='yes' matrix='1 0 0 1 219.48 456.24'> you already have A</text> <text stroke='0 0 1' pos='0 0' size='16.2' transformable='yes' matrix='1 0 0 1 370.8 456.24'>dobe Acrobat </text> <text stroke='0 0 1' pos='0 0' size='16.2' transformable='yes' matrix='1 0 0 1 67.32 437.64'>Reader i</text> <text stroke='0 0 1' pos='0 0' size='16.2' transformable='yes' matrix='1 0 0 1 131.28 437.64'>n</text> <text stroke='0 0 1' pos='0 0' size='16.2' transformable='yes' matrix='1 0 0 1 141.12 437.64'>stalled on your computer.</text> <text stroke='0 0 0' pos='0 0' size='16.2' transformable='yes' matrix='1 0 0 1 337.92 437.64'> </text> <text stroke='0 0.502 0' pos='0 0' size='12.6' transformable='yes' matrix='1 0 0 1 342.48 437.64'> </text> <image width='800' height='600' rect='-92.04 800.64 374.4 449.76' ColorSpace='DeviceRGB' BitsPerComponent='8' Filter='DCTDecode' length='369925'> feedcafebabe... </image> </page> </ipe>
We use Xpdf in one of our applications. Its a c++ library which is primarily used for pdf rendering, although it does have a text extractor which could be useful for this project.