Cannot extract tables using PdfTxtlayerReader

[ValiullinAlbert]

When I read a document using PdfTxtlayerReader, there are no tables even though the document contains some. This document is copyable. Could you tell me what I am doing wrong?

[NastyBoget]

Try to use need_pdf_table_analysis="True" in the parameters, e.g.:

reader = PdfTxtlayerReader(config={})
document = reader.read(file_path, parameters=dict(need_pdf_table_analysis="True")

If this doesn't help, could you please send us the document and the code you are using to handle it? We'll try to check it and fix as soon as possible

[NastyBoget]

There is also one notice: try to use PdfTabbyReader instead, it also works with copyable PDF documents

[ValiullinAlbert]

Thank you! table.pdf It didn't help me when I used need_pdf_table_analysis="True" in the parameters. The code:

from dedoc.readers import PdfTxtlayerReader, PdfTabbyReader

filepath = "table.pdf"
reader = PdfTxtlayerReader(config={"n_jobs": 1})
document = reader.read(filepath, parameters=dict(need_pdf_table_analysis="True"))
document.tables

Although when using PdfTabbyReader it works

[NastyBoget]

I guess the problem is that PdfTxtlayerReader extracts tables on images without using a textual layer :( As the table has colored header, table detection hasn't worked properly, and table's text moved to the document's lines :(. On the contrary, PdfTabbyReader uses a textual layer to obtain tables from PDF, and the processing was successful.

We will try to solve the problem with colored tables in the future.

Do you still have a problem?

[NastyBoget]

Here in PDF handling section is description of the parameter pdf_with_text_layer - if it's true, then PdfTxtlayerReader is used, if it's tabby - PdfTabbyReader is used. The description contains a little bit information about tables handling - PdfTxtlayerReader extracts tables as it PdfImageReader does (false option)

[ValiullinAlbert]

The problem isn't only about color headers. For example, with attached file it's still not working Table Example - Starting.pdf

[NastyBoget]

PdfTxtlayerReader can extract the attached image. We are aware of the problem with attached images in PdfTabbyReader and already solved it (on the develop branch). We are going to make a new release in the end of this week.

[ValiullinAlbert]

Sorry, I mean that attached file is my pdf file. I wanted to say that PdfTxtlayerReader cannot recognize tables without colored headers

[NastyBoget]

I'm sorry for misunderstanding, try to use config provided by dedoc (worked for me):

from dedoc.config import get_config
from dedoc.readers import PdfTxtlayerReader

reader = PdfTxtlayerReader(config=get_config())
document = reader.read("Table.Example.-.Starting.pdf")
print(len(document.tables))

We will fix this behavior in the future, thank you for noticing that

[ValiullinAlbert]

Thank you, this works!