Bring Reader into conformance with other resumable decoders - #6
Merged
Conversation
On input not ending in the row-terminating empty line, Reader._finalize used to emit the incomplete trailing row — the non-resumable behavior, diverging from the Java/Rust/Scala readers which buffer it. Align with them: _finalize no longer flushes the partial cell/row buffers, so readRow() returns null without emitting the tail. Non-resumable parse() keeps emitting the incomplete tail; existing Reader tests now use properly terminated input. Empty-row handling in _processChunk is a separate known issue and is left untouched here.
Three scaling fixes in the streaming Reader, all the same shape: batch work instead of paying per smallest unit. - _processChunk: scan chunks with indexOf and slice whole lines instead of appending to a string char by char; a line split across chunks accumulates pieces in a list joined once on completion. 16MB cell arriving in 8KB chunks: 9.6s -> 0.014s. - unescape: return early when the cell has no backslash. - readRow: consume the row queue via head index instead of shift(), which is O(n) per call once the queue is large; queue and head reset when drained. 1M-row drain went from minutes to 0.37s. Differential-tested against parse() over the nsv-tests enum+valid corpora at every chunk-split pair, and the 92MB champernowne fixed point streamed row-identical.
The row stream emits only terminated rows, and end of stream is final in Node, so without an accessor the withheld tail would be destroyed. partialRow() returns a copy of the row in progress — completed cells plus the unfinished cell run through unescape — or null, queryable at any point. For any input, readRows() plus partialRow() equals parse().
The streaming layer no longer parses incrementally: it scans each chunk for the last row boundary (a newline preceded by a newline, with one carried bit for chunk-spanning pairs and stream start), hands everything up to it to parse(), and keeps the rest as the raw pending tail. The Reader can no longer disagree with parse() on row content by construction. partialRow() becomes partial(): the raw encoded tail, empty when none. Consumers who want the truncated row as cells run it through parse() themselves. For any input and chunking, readRows() concatenated with parse(partial()) equals parse(input).
_pieces -> _tailPieces, _rowHead -> _rowQueueHead, _done -> _ended. The carried previous-byte-was-newline bit was redundant with the tail itself (empty or ending in newline), so it's a derivation now, not a member. Drop the _processChunk comment essay.
Rope concatenation makes string append O(1), so the fragment list was only earning its keep by making the line-start check readable without flattening; carrying that one bit explicitly is cheaper than the list. _partial mirrors the partial() accessor it backs. Also dedupe the _finalize doc block left by the merge and rename tests to row terms.
Since the unterminated final row stopped being flushed at end of input, _finalize was a one-line flag set with a name promising more.
Header described newlines as separators — the spec's named pitfall; they terminate. Reader class doc claimed per-chunk parsing it no longer does. partial() moved out of the middle of the read-method family. readRows/iterator no longer re-start what readRow starts.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.