Added in a retrieval function for the tables containing financial data.

[dxlnr]

Added a retrieve_html_tables function in extract_items to retrieve all the tables containing financial data as pandas dataframes. This can be used to store them in a csv or similar file format.

[eloukas]

Hi @dxlnr, thanks for the pull request! 👋
Indeed, this is something that people are interested in. Before merging it, we need to see how well it works.

Could you provide the output result for some (e.g., 4 or 5) reports of different companies' reports?
(I am asking this because tables structure might change between different years and different companies).

For example, you could copy/paste (or screenshot) the relevant table in the report, and then the code's output.
For reproduction purposes, it would be good to also say which report is that, so I can reproduce it, before merging it into the codebase.

Once again, many thanks for the contribution 🙌

[dxlnr]

Hi! I added two more commits, with minor fixes and a testing pipeline for retrieving the tables. Please find the results in the tests/fixtures/EXTRACTED_TABLES.zip. I used the AMD filings of 2017 and 2022 and kept the table extraction separate from the raw text/json extraction. I decided to create one csv file for all table from a single filing. Maybe there is a better solution for collecting these tables. (Problem is that they have different column numbers throughout).

[dxlnr]

Only issue that remains is that the table rows with blank space like the one in the image (Diluted & Basic). Gets extracted with
,,Basic,952,952,835,835
,,Diluted,"1,039",952,835,835
Which is somewhat wrong.

You would want it to be
Basic,952,,952,835,,835
Diluted,"1,039",,952,835,,835

[dxlnr]

Other than that, I created an additional function extract_tables in extract_items.py to make it testable. This function used parts of the extract_items function which might be fused in the future to avoid duplicated code.

[dxlnr]

@eloukas Can I help along with any issues to get it merged eventually?