Conversation
dajac
force-pushed
the
topic-ids-to-array
branch
from
September 18, 2026 07:57
8749b2c to
d6a3750
Compare
dajac
force-pushed
the
uniform2-util
branch
from
September 18, 2026 07:57
12cfd9a to
cde85d8
Compare
dajac
commented
Sep 18, 2026
Comment on lines
+22
to
+24
| * A minimal growable list of ints for the uniform2 assignor, which avoids the boxing of a | ||
| * {@code List<Integer>}. The assignor collects members, topics and partitions in many small | ||
| * lists, so the boxed integers and their hashing would cost more than the assignment itself. |
Owner
Author
There was a problem hiding this comment.
nit: Let's avoid the ref to the uniform2 assignor. It is a generic collection so its documentation must be generic. Also, look at the recent comments regarding your comments to ensure that you incorporate the previous feedback.
| * @param capacity The initial capacity. The list grows as needed from at least one value. | ||
| */ | ||
| public IntArrayList(int capacity) { | ||
| values = new int[Math.max(1, capacity)]; |
Owner
Author
There was a problem hiding this comment.
nit: Should it throw if the capacity is negative? Could it be zero?
| /** | ||
| * Removes the first occurrence of the value, if any, shifting the following values down. | ||
| */ | ||
| public void removeValue(int value) { |
Owner
Author
There was a problem hiding this comment.
nit: Should we name it remove to align with add?
Comment on lines
+26
to
+29
| * An immutable {@link Set} of integers backed by a sorted array without duplicates, for the | ||
| * partition sets of the assignments built by the uniform2 assignor. It is much cheaper to build | ||
| * and to store than a {@code HashSet<Integer>}, which matters when an assignment holds many | ||
| * small partition sets: building one copies a sorted array, and a lookup is a binary search. |
Owner
Author
There was a problem hiding this comment.
ditto about the documentation here.
Comment on lines
+22
to
+24
| * A minimal binary min-heap of longs for the uniform2 assignor, which avoids the boxing of a | ||
| * {@code PriorityQueue<Long>}. The assignor orders members by keys packing a load and a member | ||
| * index into one long, so the smallest key is the next member to serve. |
| * @param capacity The initial capacity. The heap grows as needed from at least one value. | ||
| */ | ||
| public LongArrayHeap(int capacity) { | ||
| values = new long[Math.max(1, capacity)]; |
Owner
Author
There was a problem hiding this comment.
same comment about the capacity.
Comment on lines
+24
to
+28
| * The dense indices of a fixed set of {@link Uuid}s, for the topics of the uniform2 assignor, | ||
| * which numbers them once and then works with the numbers. The ids are kept in open addressing | ||
| * tables of their two longs, so a lookup hashes and compares longs without boxing, which is much | ||
| * cheaper than a {@code HashMap<Uuid, Integer>} for the lookups the assignor does in large | ||
| * numbers. |
dajac
force-pushed
the
uniform2-util
branch
from
September 18, 2026 13:10
cde85d8 to
f5d91e8
Compare
The uniform2 consumer group assignor, which follows in its own change, works on members, topics and partitions numbered from zero and keeps its state in arrays. Four small collections in the package assignor.uniform2.util let it do so without boxing or hashing, which matters for the memory and the time an assignment takes on groups of thousands of members and topics: - IntArrayList, a growable list of ints, for the many small lists of members, topics and partitions the phases collect. - IntArraySet, an immutable Set<Integer> over a sorted array, for the partition sets of the assignments: cheaper to build and to store than a HashSet<Integer>, with a binary search for lookups, and equal to any Set with the same integers. - LongArrayHeap, a binary min-heap of longs, for the members ordered by keys packing a load and a member index into one long. - UuidIndex, the dense indices of a fixed set of Uuids kept in open addressing tables of their two longs, for the topics, numbered once and then referred to by number. The package is meant for the assignor only, not as a module wide toolbox. Every class and public method is documented, and unit tests cover every method, including the growth of the list and the heap, the Set contract of IntArraySet and the colliding slots of UuidIndex. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
dajac
force-pushed
the
uniform2-util
branch
from
September 18, 2026 14:26
f5d91e8 to
c23bdcf
Compare
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
1 participant
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.
This is the first of three changes adding the uniform2 consumer group assignor. It adds the four small collections the assignor is built on, in the package
assignor.uniform2.util. The assignor works on members, topics and partitions numbered from zero and keeps its state in arrays, so these collections let it avoid boxing and hashing, which matters for the memory and the time an assignment takes on groups of thousands of members and topics.IntArrayList: a growable list of ints, for the many small lists of members, topics and partitions the phases collect.IntArraySet: an immutableSet<Integer>over a sorted array, for the partition sets of the assignments. It is cheaper to build and to store than aHashSet<Integer>, looks values up by binary search, and is equal to anySetwith the same integers.LongArrayHeap: a binary min-heap of longs, for members ordered by keys packing a load and a member index into one long.UuidIndex: the dense indices of a fixed set ofUuids, kept in open addressing tables of their two longs, for the topics.The package is meant for the assignor only. Every class and public method is documented, and unit tests cover every method, including the growth of the list and the heap, the
Setcontract ofIntArraySetand the colliding slots ofUuidIndex.