[darcs-users] darcs patch: Info.lhs, OldDate.lhs: push bytestring d... (and 3 more)

Sat Apr 26 11:11:42 UTC 2008

On Sat, Apr 26, 2008 at 12:24:53AM -0400, gwern0 at gmail.com wrote:
> Thu Apr 24 20:27:39 EDT 2008  gwern0 at gmail.com
>   * Info.lhs, OldDate.lhs: push bytestring down the hierarchy
>   We don't want unpackPS/packString high up in the call hierarchy; we want those functions being called only at the very site or definition they are needed.

I'd prefer to have long comments on patches word-wrapped.

>   Unfortunately, we can't completely ByteString'ify OldDate.lhs, although I've given a sketch how to make it more bytestringy.
> 
> Sat Apr 26 00:04:47 EDT 2008  gwern0 at gmail.com
>   * FastPackedString, OldFastPackedString.hs: add in 'allPS' definition
>   This is a missing list utility function; for FPS.hs, it's the usual wrapping, while I copy the Bytestring version (swapping 'unsafePeformIO' for 'inplacePerformIO') into OldFPS. This will be a useful function for the next patch.
> 
> Sat Apr 26 00:07:57 EDT 2008  gwern0 at gmail.com
>   * OldDate.lhs: with allPS defined, we can define parseDate as more bytestringy
> 
> Sat Apr 26 00:23:24 EDT 2008  gwern0 at gmail.com
>   * refactor some of parseDate in OldDate.lhs

Content-Description: A darcs patch for your repository!
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA512
> 
> 
> New patches:
> 
> [Info.lhs, OldDate.lhs: push bytestring down the hierarchy
> gwern0 at gmail.com**20080425002739
>  We don't want unpackPS/packString high up in the call hierarchy; we want those functions being called only at the very site or definition they are needed.
>  
>  Unfortunately, we can't completely ByteString'ify OldDate.lhs, although I've given a sketch how to make it more bytestringy.
> ] hunk ./src/Darcs/Patch/Info.lhs 108

This patch has way too many white-space changes.  Please resend without
them.  I don't care to read through page after page of non-changes, in
order to seek out what (according to the patch description) should be a
very simple change.

>  --   before 1.0.  Fortunately, newer patch dates are written in
>  --   UTC, so this timezone truncation is harmless for them.
>  readPatchDate :: PackedString -> CalendarTime
> - -readPatchDate = ignoreTz . readUTCDate . unpackPS
> +readPatchDate = ignoreTz . readUTCDate
>    where ignoreTz ct = ct { ctTZ = 0 }
>  
>  pi_date :: PatchInfo -> CalendarTime
> hunk ./src/OldDate.lhs 30
>  import System.Time
>  import Data.Char ( toUpper, isDigit )
>  import Control.Monad ( liftM, liftM2 )
> +import FastPackedString
>  
>  -- | Read/interpret a date string, assuming UTC if timezone
>  --   is not specified in the string
> hunk ./src/OldDate.lhs 34
> - -readUTCDate :: String -> CalendarTime
> +readUTCDate :: PackedString -> CalendarTime
>  readUTCDate = readDate 0
>  
> hunk ./src/OldDate.lhs 37
> - -readDate :: Int -> String -> CalendarTime
> +readDate :: Int -> PackedString -> CalendarTime
>  readDate tz d =
> hunk ./src/OldDate.lhs 39
> - -             case parseDate tz d of
> +             case parseDate tz (unpackPS d) of
>               Left e -> error e
>               Right ct -> ct
>  
> hunk ./src/OldDate.lhs 43
> +-- This is hard to convert to ByteString because Parsec's 'parse' needs String
> +-- and because the only 'read' for ByteString is in the separate Data.Binary package.
>  parseDate :: Int -> String -> Either String CalendarTime
>  parseDate tz d =
>                if length d >= 14 && and (map isDigit $ take 14 d)
> hunk ./src/OldDate.lhs 117
>  cvs_date_time tz =
>                  do y <- year
>                     char '/'
> - -                   mon <- month_num 
> +                   mon <- month_num
>                     char '/'
>                     d <- day
>                     my_spaces