I am trying to convert pdf to csv file. pdf file has data in tabular format with first row as header. I have reached to the level where I can extract text from a cell, compare the baseline of text in table and detect newline but I need to compare table borders to detect start of table. I do not know how to detect and compare lines in PDF. Can anyone help me?
Thanks!!!
As you've seen (hopefully), PDFs have no concept of tables, just text placed at specific locations and lines drawn around them. There is no internal relationship between the text and the lines. This is very important to understand.
Knowing this, if all of the cells have enough padding you can look for gaps between characters that are large enough such as the width of 3 or more spaces. If the cells don't have enough spacing this will unfortunately probably break.
You could also look at every line in the PDF and try to figure out what represents your "table-like" lines. See this answer for how to walk every token on a page to see what's being drawn.
I was also searching the answer for the similar question, but unfortunately I didn't found one so I did it on my own.
Here is the github link for the dotnet Console Application I made. https://github.com/Justabhi96/Detect_And_Extract_Table_From_Pdf
This application detects the table in the specific page of the PDF and prints them in a table format on the console. Here is the code that i used to make this application.
First of all I took the text out of PDF along with their coordinates using a class which extends iTextSharp.text.pdf.parser.LocationTextExtractionStrategy class of iTextSharp. The Code is as follows:
This is the Class that is going to store the chunks with there coordinates and text.
And this is the class that extends the LocationTextExtractionStrategy class.
This class is overriding the RenderText method of the LocationTextExtractionStrategy class which will be called each time you extract the chunks from a PDF page using PdfTextExtractor.GetTextFromPage() method.
Now for Detecting the start and end of the table you can use the coordinates of the chunks extracted from the PDF. Like if the specific line is not having table then there will be no jumps in the right coordinate of the current chunk and and Left coordinate of next chunk. But the lines having table will be having those coordinate jumps of at least 3 points.
Like for Lines having table will have coordinates of chunks something like this:
right coord of current chunk -> 12.75pts
left coords of next chunk -> 20.30pts
so further you can use this logic to detect tables in the PDF. The code is as follows:
You can call getLineText() method of the above class and run the following loop to see the output in the table structure on the console.