I have a pdf of which the content stream of the pdf doc looks like image1.
But once I open the pdf in adobe dc and tried to change the reading order. The entire content stream is changed. (Please see image2)
And here is the link to source pdf https://drive.google.com/file/d/1V2K3-2GdWG5DuTUv1fyfIIT54en70kI2/view
Is there a way to do the same programmatically(convert content stream of graphical text to proper stream)
Thanks in advance !
Is there a way to do the same programmatically(convert content stream of graphical text to proper stream)
First of all, both streams are proper, there merely are different (and in the case at hand considerably different) ways to create the same text on screen, each of them as valid as each other, and different PDF processors use different ways.
The processor that created your original PDF appears to have approached the task by dividing the text in small pieces (less than a text line) and draw these pieces as independently as possible, i.e. as separate text objects (BT..ET) with text properties set in each (Tm, Tf, Tc), positioned and rescaled by transformation matrix changes (cm), enveloped in save/restore graphics state instructions (q..Q).
Adobe Acrobat, on the other hand, appears to prefer the page main text to be contained in a single text object with text properties only set when they change and no text object or graphics state switches in-between.
Neither of these is more "proper" or more "graphical" than the other. If anything, these structures mirror how these instructions are stored or processed internally by the respective PDF processor.
That being said, you do want to convert from the former style into the latter.
The main problem is that the latter style is not standardized (at least there is no published document normatively describing it). So, while you can surely attempt to follow the lead of the example you have, you can never be sure that you understood the style exactly. Thus, you always have to expect differences emerging in special, not yet encountered situations. Furthermore, there is no guarantee Adobe will meticulously adhere to that style across software versions.
Nonetheless, you can of course attempt to follow the style (as you perceive it) as well as possible.
An implementation will have to walk through the respective content stream, keeping track of the current graphics state, and transform the text drawing (and related) instructions into a single text object for as long as possible.
You have tagged your question both itext and pdfbox. Thus, you appear to be undecided with which PDF library to implement this. Here some ideas for both choices:
For processing content streams and keeping track of the current graphics state, iText offers its com.itextpdf.text.pdf.parser API, in particular the PdfContentStreamProcessor (iText 5.x) / its com.itextpdf.kernel.pdf.canvas.parser API, in particular the PdfCanvasProcessor (iText 7.x).
You can extend them to in addition to analyzing the current contents also replace the content stream in question with an updated version, e.g. like I did in this answer for iText 5 or in this answer for iText 7.
PDFBox for the same task offers the class hierarchy based on the PDFStreamEngine. Based on these classes it should similarly be possible to create a graphics state aware content stream editor.
Both libraries also offer simpler classes for parsing the content streams into sequences of instructions, but those classes don't keep track of the graphics state, leaving that for you to implement.
