Parsing program config files

Question

I need to write reader class for a program config file, and I'm trying to determine the best way to approach this.

I have some experience with boost::spirit (actually, just enough to know that I am terrible at writing grammars), and a good bit of experience with regular-expressions.

Other than that, I can't think of much. So my options seem to be:

boost::spirit
std or boost regex
ol' fashion string parsing

I am looking for suggestions or advice on how to proceed - or even pseudo-code should anyone feel inclined.

I realize this is a bit of an opinion-based question, but I'm really struggling with writers-block here. I'm worried I'll start down a path, and waste a lot of time before realizing it was a poor choice.

The config file format is already defined, and looks like this:

Group1 {
    Scalar1 = 500
    Scalar2 = 45.5
    Scalar3 = My Value
    List1 {
        LS1 = 123
        LS2 = hello world
    }
    List2 {
        LS1 = 456
        LS2 = goodbye world
    }
    Array1 [
        300
        200
        25
    ]
    Array2 [
        true
        false
        true
        false
    ]
};

The config will always have at least one Group.
A Group will contain 0..n Scalar, List, and/or Array entries
A scalar is: label = value
A List is a labeled container of Scalars
An Array is a labeled container of values

The labels shown (eg. Scalar1, Scalar2, ...) are just examples. They can be called anything: maxrate = 500, avg_val = 45.5, would have been just as valid.

quick question: why do you want to define your own config format when there are already quite a few config formats out there? — DZDomi, Jun 22 '18 at 00:15
@DZDomi Thanks for the comment. The config format is already defined. I need to write a lib to conform to the format that is already in use. I didn't make that clear. If you have a suggestion, though, I could propose changing the format as a possible path. — Blair Fonville, Jun 22 '18 at 00:17
The language looks small. IMHO, unless you have extensive experience with Boost::Spirit, Flex or Bison, write your own. I could write this up faster than I could get the above tools working. I recommend in any case, draw a parsing diagram and maybe also BNF format. — Thomas Matthews, Jun 22 '18 at 00:20
@ThomasMatthews beat me to it. Only thing I'd add is I don't think Regex is a good fit for this one. — user4581301, Jun 22 '18 at 00:21
Look up recursive descent parsing, this would be very easy and clean done that way. Unless you have a grammar for numeric literals I'd just define them as "whatever strtod and strtol parse" — gct, Jun 22 '18 at 00:23
@ThomasMatthews I was thinking the same. My biggest concern though, are the scalar values, which can be strings. And the strings could contain []{} characters. This would complicate things a little as I'd have to be careful to match openings and closings. I didn't know if a grammar would be better suited for that reason. — Blair Fonville, Jun 22 '18 at 00:26
@SeanMcAllister Not sure how I missed that. It's perfect for recursive descent - and I already have some code I can reuse. Thanks — Blair Fonville, Jun 22 '18 at 00:31
@BlairFonville Glad to hear it. For strings and such, I'd write a regexp to try to suck them out. — gct, Jun 22 '18 at 00:32
I never said to not develop a grammar. Some grammars are not worth the hassle of the big tools. — Thomas Matthews, Jun 22 '18 at 02:05
@ThomasMatthews No, you’re right. In fact you specifically said that I should. I misspoke - I meant to write “I didn’t know if a grammar library would better suited”. If I could trust that there would be no human error, the parsing would be trivial. My primary goal is robust error detection and handling. That’s where it gets most complicated. I probably should have stated that in the question. — Blair Fonville, Jun 22 '18 at 02:24
See my article on how to write a recursive descent parser. This is actually pretty easy. https://stackoverflow.com/questions/2245962/is-there-an-alternative-for-flex-bison-that-is-usable-on-8-bit-embedded-systems/2336769#2336769 — Ira Baxter, Jun 22 '18 at 09:11

score 1 · Accepted Answer · answered Jun 22 '18 at 06:02

1

Copy the file to tmp, replacing every instance of = with :, then read it as json with boost::ptree.

answered Jun 22 '18 at 06:02

Red.Wave

2,790
11
17

While this doesn't work "out of the box", I can definitely work with this, and it's probably the quickest path to take. It's just what I was looking for. Thanks. – Blair Fonville Jun 25 '18 at 19:06
you're wellcome. The config syntax was really similar to json. Actually if the assignment were replaced with colon, it'd be json. – Red.Wave Jun 25 '18 at 20:08
I also had to add commas after each entry, and wrap entry names and string values (where values included spaces) in quotes. But it was a quick fix, and gave me the error detection that I needed. – Blair Fonville Jun 25 '18 at 20:16

Parsing program config files

1 Answers1