There are some INI files where the order of the sections is relevant.
Look for example at the pacman.conf of the Arch Linux package manager (which is my use case).
I don't think giving up on the HashMap as an underlying datastructure is the right way since it's sufficient for probably most of the use cases, but would an order-preserving version in a separate module be acceptable? Or a parameterized Ini type and Ini simply be a type alias like

newtype Ini' f = Ini' { unIni :: f Text (HashMap Text Text) }
type Ini = Ini' HashMap

If there's any chance such an extension were accepted I'd volunteer for the implementation.
Thanks!

Nothing bad, just location: http://github.com/chrisdone/inil.git yields a 404.

I had expected that, if I read a completely empty file with readIniFile, it would return Ini (fromList []). Instead, the parser fails with "not enough input".

A similar problem occurs with a file that contains an empty section.

ghci> writeIniFile "foo.ini" $ Ini (fromList [])
ghci> readIniFile "foo.ini"
Left "not enough input"
ghci> writeIniFile "foo.ini" $ Ini (fromList [("foo", empty)])
ghci> readIniFile "foo.ini"
Left "not enough input"

Sorry for the lack of narrowing this down, but I didn't want to just let things pass without letting you know about this issue. If you want to flag this as an unsupported format and close it, that'd be fine by me.

I'm trying to parse ini files from deviationTx (deviationTx.com), which uses some odd features of the ini file format. Tried feeding one to readIniFile to see if I could use this Data.Ini, and got back an empty list (Right (Ini {unIni = fromList []}).

Hi. Thanks for writing this library.
This is not as much of an issue as it is a question - how does one edit the parsed ini file? I need to replace some values and put the data in the file it was read from. Iterating over the structure to find 2 changes and write it back, line-by-line seems over the top. Newbie to haskell so I am not sure if functionality of editing the ini file is even something that can be done.

Thanks.

Using the same INI content as is in the documentation (see attachment), I am able to get an empty return value from readIniFile simply by inserting a single empty line in front of the ini content.

Presumably the package should be able to handle insignificant empty lines?

test.ini.zip

0.3.6 loads Data.Semigroup without specifying semigroups as a dependency in its cabal file, nor specifying a base dependency for ghc 8 or greater which would cause the resolver to seek a working version automatically.

This means that people on 7.10 and lower can't use the package easily.

If I may, I would suggest adding

  if !impl(ghc >= 8)
    build-depends: semigroups >= 0.10 && < 0.19

to your cabal file.

When semigroupally combining Ini values that both define the same section, I was surprised to find that one of the sections is discarded. The monoid instance currently looks like this:

mappend x y = Ini {iniGlobals = mempty, iniSections = iniSections x <> iniSections y}

I was expecting sections of the same name to be merged.

Also, why discard all the globals?

Proposed change to the monoid instance:

mappend x y = Ini {
    iniGlobals = Map.unionWith (<>) (iniGlobals x) (iniGlobals y),
    iniSections = Map.unionWith (<>) (iniSections x) (iniSections y) }

Either way, whether you are interested in changing the instance or not, I'd be happy to send a PR with some documentation to mention how the instance behaves.

Repo here: https://github.com/cies/htoml

Would you mind uploading a new version to Hackage?

Due to this change only GHC 7.10.* can new be used.
Not sure if GHC 7.8.* is supposed to be supported though.