Hi Peter,

Am 7/2/2013 12:52 AM, schrieb Peter Hall:

The other problem I'm facing is with name collisions for record
accessors. There are two versions of MtGoxTicker from different services
- one with a subset of the properties, which is supposed to be faster
(it isn't, but that isn't the point). The data types are here
https://github.com/peterjoel/auto-trader/blob/6974d66ae51459479c19be291d075bbdeb718b53/AutoTrader/MtGox/Types.hs.
One is commented one out while I decide what to do. What is the best way
to model those records to avoid collisions, while not being confusing to
users of the library? I am very tempted to use type classes, but that
feels naughty. Using unique prefixes seems bad too - it would be nice
for some code to be able to use them interchangeably if they don't need
all the fields.

This sounds like a good case for two separate modules:

AutoTrader.MtGox.Ticker.Full would have a data Ticker = Ticker { .. } which is the commented-out MtGoxTickerFull, and AutoTrader.MtGox.Ticker.Fast would balso have a 'data Ticker = ...', which would be your MtGoxTicker.

This would allow you to use the same field names without getting clashes, users of your library could choose hwo to import the types, i.e. what prefix to use - and people using no fields from the full ticker could switch their code by changing something like

  import qualified AutoTrader.MtGox.Ticker.Full

to

  import qualified AutoTrader.MtGox.Ticker.Fast

As a side note, I think that in the vast majority of cases where you think that a type class is be a good solution - it's not. :-}

--
Frerich Raabe - raabe@froglogic.com
www.froglogic.com - Multi-Platform GUI Testing


_______________________________________________
Beginners mailing list
Beginners@haskell.org
http://www.haskell.org/mailman/listinfo/beginners