allow warming partially an sstable for an automaton #2559
Merged
Conversation
This file contains 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
fulmicoton
reviewed
Dec 19, 2024
fulmicoton
reviewed
Dec 19, 2024
fulmicoton
reviewed
Dec 19, 2024
fulmicoton
reviewed
Dec 19, 2024
fulmicoton
reviewed
Dec 19, 2024
fulmicoton
reviewed
Dec 19, 2024
fulmicoton
reviewed
Dec 19, 2024
fulmicoton
reviewed
Dec 19, 2024
fulmicoton
reviewed
Dec 23, 2024
fulmicoton
reviewed
Dec 23, 2024
fulmicoton
approved these changes
Dec 23, 2024
fulmicoton
reviewed
Dec 26, 2024
fulmicoton
reviewed
Dec 26, 2024
fulmicoton
reviewed
Dec 26, 2024
src/index/inverted_index_reader.rs
Outdated
| // TODO this operation is often cheap for "friendly" automatons, but can be very costly for | ||
| // "unfriendly" ones such as ".*a{50}" (very few terms if any match this pattern, but we | ||
| // can't know early). In this case, we decompress and iterate over the entire sstable, while | ||
| // still being in async context. Ideally we should spawn this on a threadpool. | ||
| let range_to_load = term_info |
There was a problem hiding this comment.
Same as above, it is difficult for readers to know what is the type of range_to_load.
fulmicoton
reviewed
Dec 26, 2024
src/index/inverted_index_reader.rs
Outdated
| // TODO this operation is often cheap for "friendly" automatons, but can be very costly for | ||
| // "unfriendly" ones such as ".*a{50}" (very few terms if any match this pattern, but we | ||
| // can't know early). In this case, we decompress and iterate over the entire sstable, while | ||
| // still being in async context. Ideally we should spawn this on a threadpool. |
There was a problem hiding this comment.
I see. So if I understand correctly, the trouble is that the computation is done within an iterator.
We eventually yield, but that's only when we effectively get an element.
Is decompressing a single block on the runtime is fast enough, we could probably write ugly code and have our iterator return Option<Range<usize>> to offer the runtime to yield more often.
If we do it this way, we might also want to actively yield.
I believe tokio only force yield on await, after await has been reach N times. (to be checked). That N might be way too high for us.
The second is probably not of a problem
fulmicoton
reviewed
Jan 2, 2025
src/index/inverted_index_reader.rs
Outdated
| .get_term_range_async(.., automaton.clone(), None, merge_holes_under_bytes) | ||
| .await?; | ||
|
|
||
| // we build a 2nd iterator, this one with no holes, so we don't go through blocks we can't | ||
| // match, and just download them to reduce our query count. This makes the assumption | ||
| // there is a caching layer below, which might not always be true, but is in Quickwit. | ||
| let term_info = self.get_term_range_async(.., automaton, None, 0).await?; | ||
| let term_infos = self.get_term_range_async(.., automaton, None, 0).await?; |
There was a problem hiding this comment.
Suggested change
| let term_infos = self.get_term_range_async(.., automaton, None, 0).await?; | |
| let term_info_stream = self.get_term_range_async(.., automaton, None, 0).await?; |
fulmicoton
approved these changes
Jan 2, 2025
|
(approved but we need to fix the runtime footgun you identified of course) |
fulmicoton
reviewed
Jan 3, 2025
src/index/inverted_index_reader.rs
Outdated
| let merge_holes_under_bytes = (80 * 1024 * 1024 * 50) / 1000; | ||
| // we build a first iterator to download everything. Simply calling the function already | ||
| // loads everything, but doesn't start iterating over the sstable. | ||
| let mut _term_infos = self |
There was a problem hiding this comment.
Suggested change
| let mut _term_infos = self | |
| let _term_infos = self |
trinity-1686a
commented
Jan 6, 2025
fulmicoton
reviewed
Jan 8, 2025
src/index/inverted_index_reader.rs
Outdated
| // we could do without an iterator, but this allows us access to coalesce which simplify | ||
| // things | ||
| let posting_ranges_iter = | ||
| std::iter::from_fn(move || stream.next().map(|(_k, v)| v.postings_range.clone())) |
There was a problem hiding this comment.
can we break it into several statement.
For instance:
- Create the iterator of postings ranges.
- doe the coalesce thing after.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.