wren:
Bryan O'Sullivan wrote:
As a case in point, I took the string search benchmark that Daniel shared on Friday, and boiled it down to a simple test case: how long does it take to read a 31MB file?
GNU wc -m:
- en_US.UTF-8: 0.701s
text 0.7.1.0:
- lazy text: 1.959s - strict text: 3.527s
darcs HEAD:
- lazy text: 0.749s - strict text: 0.927s
When should we expect to see the HEAD stamped and numbered? After some of the recent benchmark dueling re web frameworks, I know Text got a bad rap compared to ByteString. It'd be good to stop the FUD early. Repeating the above in the announcement should help a lot.
For what its worth, for several bytestring announcements I published comprehensive function-by-function comparisions of performance on enormous data sets, until there was unambiguous evidence bytestring was faster than List. E.g http://www.mail-archive.com/haskell@haskell.org/msg18596.html