On 26 Sep 2009, at 1:40am, BareFeet wrote: > As above, I think it is pretty simple. The complexity arrives when > people expect CSV to provide more than a simple array of strings.
One of my free programs has to read CSV files ( http://www.hearsay.demon.co.uk/mac/MacGarminTools/csv2gpx.html if you care). The documentation for the program doesn't mention any of the bizarre things people expect to work when they use it. For about a year after I released the program I got the occasional email from people who had been trying to use it with a CSV file which JUST SHOULD WORK, DAMMIT. They had all sorts of bizarre things in their files, from methods of escaping characters which I'd never heard of, including return characters in fields (For describing GPS points ? You really need more than one line of text ?), and marking header and comment lines in all sorts of weird ways. Quite a lot of them could point to some big data publisher which was using their weird format as if it was well-known. Also some files are in Latin 1 and some are in UTF-8. Consequently, my software deals with all sorts of weird things in CSV files and very few of them are mentioned in the documentation. I have no doubt that much commercial software does the same thing: contains tweaks wanted by one user which would be too ridiculous to document and create too many opportunities for requests for backward compatibility. I'm fortunate that that software is seldom used these days, now that the manufacturer supplies software which performs that function. I suspect that three years later I'd still be getting requests to include a new CSV dodge. Simon. _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users