Write tomland benchmarks

[chshersh]

We're interested only in bechmarks for parsing (mostly because other Haskell TOML libraries don't have pretty-printing abilities) and for converting to Haskell data types (aka 2 phases). See this talk by George Wilson about CSV encoding/decoding library for multiple phases explanation:

• https://www.youtube.com/watch?v=zP-DgjIcCho

I would like to have a comparison with the following libraries:

• htoml
• htoml-megaparsec
• toml-parser *conflicted in deps
• toml *conflicted in deps

Also, I recommend using gauge library for the benchmarks.

• Add function with type like TomlCodec a -> TOML -> Either DecodeException a #109 Add function with type like TomlCodec a -> TOML -> Either DecodeException a
• Example of TOML for the benchmarks #110 Two easy tasks:
  • Implement example of TOML for the benchmarks.
  • Write data type, that describes this example.
• Write parser and converter for every library separately #111 Write parser and converter for every library separately.

[willbasky]

What functions/actions/operations to benchmark?
Parsing and converting middle state and converting to Haskell type. I see.

Do all other libraries have two step converting?

Does tomland has separate function for parsing and for converting?
decode :: TomlCodec a -> Text -> Either DecodeException a function includes both steps.

[willbasky]

htoml has parser step.
parseTomlDoc :: String -> Text -> Either ParseError Table where Table is HashMap with Text keys and Node as value, the last is like AnyValue in tomland.

parseOnly :: Parsec Text (Set [Text]) a -> Text -> Either ParseError a is like tomland's decode.
Thereby htoml has two steps as tomland. Is it correct?

htoml-megaparsec has the same approach.

[willbasky]

toml-parser and toml have just parser, without convertor. Is it correct?

[chshersh]

@willbasky

What functions/actions/operations to benchmark?

I suppose we should have some TOML example file with multiple fields:

• Bool
• Integer
• Text
• DateTime
• Array
• Table1
• Table2

I would like to have different types of TOML values, so we can put inside tables 1 and 2 something different. For example, if you put ZonedTIme on top-level DateTime then you can put LocalDay in table1 and just time in Table2. Same for array: you can have array of strings, array of different doubles (with nan and inf) and array of array of integers. String can be either raw or quoted or multiline string. Just some random different values to test everything.

Do all other libraries have two step converting?

Not sure. But for those, who don't have, we need to write conversion manually by our own just for this library.

Does tomland has separate function for parsing and for converting?
decode :: TomlCodec a -> Text -> Either DecodeException a function includes both steps.

tomland has separate function only for parsing:

• http://hackage.haskell.org/package/tomland-0.4.0/docs/Toml-Parser.html#v:parse

But it doesn't have separate function only for conversion. Probably it's a good idea to have one.

So I I see the following steps to work on this issue:

1. Add function with type like TomlCodec a -> TOML -> Either DecodeException a
2. Implement example of TOML for the benchmarks.
3. Write data type, that describes this example.
4. Write parser and converter for every library separately.
5. Combine everything with benchmarks.

I think that for every task you can create separate issue to make amount of work per issue smaller.

[willbasky]

• Add function with type like TomlCodec a -> TOML -> Either DecodeException a #109 Add function with type like TomlCodec a -> TOML -> Either DecodeException a
• Example of TOML for the benchmarks #110 Two easy tasks:
  • Implement example of TOML for the benchmarks.
  • Write data type, that describes this example.
• Write parser and converter for every library separately #111 Write parser and converter for every library separately.