least error-prone C scanning

[ielo]

Hi,

I often use C for data handling, with the data being sent back and forth from someone else's parent program (so I rarely write my own read/write functions). I have to update what I consider an error-prone way of scanning and parsing a given file (known format).

I've gone through a number of posts in this forum on scanning/parsing text files with C functions. Usually, a newbie posts some first-try, and then others suggest good and not so good changes to the initial code. Sometimes, the low-level char-by-char manipulation seems to be one of the less error-prone ways of doing it.

I'd like to tackle this from another angle: has anyone come across well-written published C code that scans/parses and that has prompted one to say: "Now, that's the way to do it! Code to emulate for a newbie!". What I mean by 'published' is that it's actually found in a distributed program. Of course, anything open source GPL (most of Linux) would be appropriate as an example since the source code is accessible.

 

[xwb]

What type of data are you parsing? Is it free form numbers, fixed format stuff, letter sequences, xml etc?

The proper way to write a parser is to first create a scanner (lexical analyser). This is a simple state machine. What tools like lex do is to create the state machine for you. Pass the output of the scanner to a syntax analyzer to get your data. Of course if you just want the tokens, you can just stop at the scanner stage.

 

[adholioshake]

Using fgets() to get lines of file into buffer, and then using sscanf() to search the buffer is less error prone than just using fscanf(). But I suppose it does depend on what data you want to pass, as xwb pointed out...


Don't hate the player - hate the game