Add the ability to extract and check tables in pdf files

[VitalyShalaev]

Many pdf files contain tables, but the pdf-test only extracts text from pages. I would like to check the contents of the tables in a more convenient way.

[asolntsev]

@VitalyShalaev Thank you for the suggestion. How exactly do you want to find the tables?

[VitalyShalaev]

@asolntsev
So far there are 2 options:

1. Automatic detection on the page and output of the result in the form of a list of lists.
2. Definition of the table by the specified column names.

[VitalyShalaev]

@asolntsev
For case 1:
Found the SpreadsheetExtractionAlgorithm at https://mvnrepository.com/artifact/technology.tabula/tabula/1.0.4.
But there are some problems with defining the text in the cells of some tables, trying to determine the reason, but not sure if I can do it.

[VitalyShalaev]

@asolntsev
For case 1:
I found a library that works well https://mvnrepository.com/artifact/e-iceblue/spire.pdf/10.1.3
Example: https://medium.com/@alice.yang_10652/extract-table-data-from-pdf-in-java-8dc4fad73752

[asolntsev]

Some thoughts from a contributor's perspective, in case it helps anyone picking this up:

The hard part of this request is that PDF has no notion of a table — it's just glyphs at coordinates, sometimes with ruling lines, sometimes not. Apache PDFBox (which this library already uses) exposes positioned text via PDFTextStripperByArea / TextPosition, but turning that into rows and columns is a heuristic problem. The mature Java option is tabula-java, but it's a fairly heavy dependency for a library whose current charm is being tiny (~3 classes).

A possible incremental path that avoids committing to a full table model up front:

1. Region-based text extraction as a primitive — e.g. pdf.textInRegion(page, x, y, w, h) backed by PDFTextStripperByArea, with matching Hamcrest + AssertJ assertions. This covers the common "assert this part of the page contains X" case without inventing a table API.
2. Table assertions on top, once there are real-world fixtures driving the API shape (header detection? whitespace per cell? row/column addressing vs. containsRow(...)?). Possibly via tabula-java, possibly via a simpler in-house heuristic if the use cases are narrow.

Happy to prototype step 1 if there's interest.