How to open PDF raw?

I've been wanting to see the inside开发者_StackOverflow中文版s of a PDF for a while, like, the raw source code of it so I can look at it. Any way of doing that?

Looking at the raw code of PDFs will not serve you much unless you also have an idea about its internal structure. You should get yourself a copy of the official PDF reference (download PDF), and you should have read some introductionary article such as this [gone] or this to begin with.

Even after such a preparation, you'll not discover much useful when staring at the raw code. Because PDFs usually will contain parts which are "filtered" (that means: compressed).

How to look at the real PDF source behind the 'raw' binary parts

Jay Birkenbilt's qpdf is a very useful commandline tool (available for Linux, Mac OSX, Windows, and as source code, under the open source Artistic License), which can unpack most filtered content and re-organize the internal structure in a way that gives you much more insight into it (all objects are numerically ordered, etc.). The commandline to achieve this is:

 qpdf  --qdf  original.pdf  unpacked.pdf

Another useful and free tool (GPL licensed, but Linux-only AFAIK) to look into PDFs is of course PDFEdit. This one even comes with a GUI (if you prefer that), while still allowing you access to the internal structure and "raw" PDF code.

If the purpose is just to look into the file, then any simple text editor will do, ex, Notepad. PDF is just a text based format, including embedded content byte streams. Raw PDF looks like this:

>>
/Border [0 0 0]
/Rect [121.02 332.48 363.24 343.64]
/StructParent 1321
/Subtype /Link
/Type /Annot
>>
endobj
64579 0 obj
<<
/Filter /FlateDecode
/Length 5771
>>
stream
Ũn0x/�+�}�ǹ����\֛ bYO�5[��X��W��L��(�������V�A3�C���������u큋_�a��ךm2N�6�    ��A��8
�d���NQ⺢GI��G�[��)�̉Y��R�y{R����&�&�;��g�k1���ҋeTC�(W��`���*��(;�AEc<=  mnZ+��|T��v
�.��зe�aޞ��V4�b���L����k�Oj.ֿ�y�����kc|I��  ��C�0��Hf�7d�/�z���m��o��A��B��IJ�%�. 
!�%f�б���&�ޒ�4Ύ7�l�3���3`�
endstream
endobj
64580 0 obj
<<
/Border [0 0 0]
/Dest <E4AE7DD2769553EF1668>
/Rect [219 648.5 256.8 659.66]
/StructParent 1323
/Subtype /Link
/Type /Annot
>>

What you see are basic COS objects like name, dictionary, stream and so on. All objects are described in PDF 32000 standard, see section 7.3 Objects.

Use a Hex editor. Of course, unless you know the PDF specification (PDF, 8.6 MB), you won't recognize much.

In addition to the qpdf tool conversion into postscript might be helpful. PDF is a subset of PS. Usually its quite easy to figure out, e.g. where the labels of a graph are. You can either use pdf2ps or invoke ghostscript

gs -sDEVICE=pswrite some.pdf -sOutputFile=some.ps -dNOPAUSE -c quit

When you generate your PDFs using pdflatex you can disable compression with an option. This makes the PDF more readable.

Some more recent observations on the other answers.

Adobe keep moving about their Open Sourced copy of the 2008 standard so currently that is here https://opensource.adobe.com/dc-acrobat-sdk-docs/standards/pdfstandards/pdf/PDF32000_2008.pdf
The Web Archive have currently a copy here https://ia601003.us.archive.org/5/items/pdf320002008/PDF32000_2008.pdf

They should be identical 22,491,828 bytes so beware neither includes any errata.

A pdf CAN be plain mime "text/pdf" as perfectly ? annotated generated from a console keyboard or command line (too slow) or a batch file. I wont bore you with the whole file but it starts like this:-

REM Start with File "Magic" Signatures for a PDF
echo %%PDF-1.0>!Fname!
echo %%âãÏÓ>>!Fname!

echo %%01) Prepare file references>>!Fname!
for %%Z in (!Fname!) do set "FZ1=%%~zZ"
echo 1 0 obj>>!Fname!
echo ^<^</Names^<^</Dests 2 0 R^>^>/Outlines 3 0 R>>/PageLayout/OneColumn/PageMode/UseOutlines>>!Fname!

REM ToDo add files
REM /Lang (ga-IE)/MarkInfo^<^</Marked true^>^>/Names ^<^<^/EmbeddedFiles [(file.ext) 3 0 R]^>^>>>!Fname!

echo /Pages 4 0 R/Type/Catalog/ViewerPreferences^<^</DisplayDocTitle true^>^>^>^>>>!Fname!
echo endobj>>!Fname!

echo %%02) Prepare Named Destinations>>!Fname!

Thus the annotated RAW PDF (note I had edited the order in the cmd file in preparation for an XMP data section, so not identical) could looks like :-

%PDF-1.3 
%âãÏÓ
%01) Prepare file references
1 0 obj
<</Lang(ga-IE)/Names<</Dests 3 0 R>>/Outlines 4 0 R/PageLayout/OneColumn/PageMode/UseOutlines
/PageLabels<</Nums[0<</S/A>>]>>/Pages 5 0 R/Type/Catalog/ViewerPreferences<</DisplayDocTitle true>>>>
endobj
%02) Reserved for big meta data
2 0 obj
<< >>
endobj
%03) Prepare Named Destinations
3 0 obj
<</Names [(Page1) [6 0 R /XYZ 0 792 null] (QRCode) [6 0 R /XYZ 25.0 317.0 1]]>>
endobj
%04) Prepare Outline / Bookmarks
...
...

Many suggestions by others for decompress binary application/PDF into text/PDF and some may be a hybrid thus still have binarized application text.

The 3 most common designed for the task are qpdf (already mentioned, but uses a hybrid QDF) PDFtk (uncompress) and Mutool (different CLI options), that's the one I play with most, as its easy in GL GUI to change the output settings. The output can be modified in MS Notepad, whilst previewing result.

So any text editing script can write or edit a PDF even with graphics, And several applications can convert RAW "binary" PDF into RAW "textual" PDF. However never attempt to edit PDF whilst temporarily in its textual base64 RePrEx (possible, but totally impractical)