Harendra Kumar
Harendra Kumar
Like streamly-bytestring. For converting text to unpinned array and vice-versa.
Stream based split operations are much faster compared to fold based. Stream ops: ``` All.Data.Stream/o-1-space.FileSplitSeqUtf8.S.splitOnSeqUtf8 abcdefghijklmnopqrstuvwxyz (1/10) 173741.00 All.Data.Stream/o-1-space.FileSplitSeqUtf8.splitOnSeqUtf8 abcdefgh (1/10) 187702.00 All.Data.Stream/o-1-space.FileSplitSeq.splitOnSuffixSeq suffix lf 94668.50 All.Data.Stream/o-1-space.FileSplitSeq.splitOnSuffixSeq suffix empty pattern...
This is an experiment, turns out to be slower than not using openat.
logs: https://github.com/composewell/unicode-transforms/actions/runs/13150321571/job/36696509226?pr=100 Anyone familiar with this CI please take a look. @adithyaov @wismill .
See also: https://github.com/composewell/streamly/issues/3039 for a small enhancement.
We have combinators with monadically generated scans which is very powerful but can be inconvenient to use when we just need some static scans. Currently we have: ``` parDistributeScan ::...
Now we represent backtracking by negative offsets in the Partial/Continue constructors. In some parsers (e.g. deintercalate) we maintain a positive count and then for backtracking we negate it when returning...
It needs to be broken into smaller modules.