There are lots of ways to do it if I want to count characters for each of the tables I need to deal with. What I'd like is something that looks at the table and uses heuristics to figure out the column widths and names. For all the tables I'm dealing with in the first instance the tables are machine generated so the columns are unlikely to change within a table, but they do change between tables.

I wrote something similar for PDF once, and also wrote Data::TableReader, but I never got around to making PDF into one of the decoders. For PDF, it made sense to look at start X addresses for segments of text, and identify it as a column if there were roughly as many text fragments starting at an X as there are estimated number of lines. Text has less granularity, so I think if I were going to try writing it for text, I would iterate lines of text and make a history of which columns have a vertical run of whitespace, and at the EOF or first blank line, see which runs of whitespace lasted from the first to the last line. Concatenate adjacent whitespace columns, and then report the space inbetween as the data columns.

It would be really awesome if you wanted to contribute a Decoder for Data::TableReader :-)

Could you please show an example how to parse the OP's table?

I find this example particularly challenging, since

• it has nested columns
• multiple subdivided head captions
• especially "Longitude" is overlapping the "empty column" limiting its data entries below.

Cheers Rolf
_{(addicted to the 𐍀𐌴𐍂𐌻 Programming Language :)
Wikisyntax for the Monastery}