pnoltes opened a new pull request, #692:
URL: https://github.com/apache/celix/pull/692
# Intro
This PR updates the Apache Celix filter to utilize typed properties entries
and also refactors the filter implementation.
My intention was to create a small PR allowing filters to work with typed
properties before starting on array support. However, I observed several issues
with the filter implementation that I also wanted to address.
# Benchmark
With this PR, a filter match will take into account the underlying
properties entry type and use it for a more efficient comparison. A Filter
match benchmark is introduced (and run before using the typed properties
entries in filter match) to ensure the desired effect.
Here are some before and after results (measured on my machine), where the
range indicates the size of the properties set used within a filter match.
## Before Refactor
```
--------------------------------------------------------------------------------------------------------------------------------------
Benchmark
Time CPU Iterations UserCounters...
--------------------------------------------------------------------------------------------------------------------------------------
FilterBenchmark_testStringCFilter/10000/process_time/real_time
14.5 ns 14.5 ns 48853882
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=68.9608M/s
FilterBenchmark_testStringFilter/10000/process_time/real_time
14.4 ns 14.4 ns 48922109
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=69.5677M/s
FilterBenchmark_testLongFilter/10000/process_time/real_time
22.3 ns 22.3 ns 31418137
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=44.8544M/s
FilterBenchmark_testBoolFilter/10000/process_time/real_time
16.3 ns 16.3 ns 42884878
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=61.21M/s
FilterBenchmark_testVersionFilter/10000/process_time/real_time
131 ns 131 ns 5332730
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=7.6231M/s
FilterBenchmark_testVersionWithQualifierFilter/10000/process_time/real_time
163 ns 163 ns 4351054
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=6.13215M/s
FilterBenchmark_testStringGreaterEqualFilter/10000/process_time/real_time
14.4 ns 14.4 ns 48966825
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=69.4911M/s
FilterBenchmark_testLongGreaterEqualFilter/10000/process_time/real_time
22.1 ns 22.1 ns 29620437
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=45.307M/s
FilterBenchmark_testDoubleGreaterEqualFilter/10000/process_time/real_time
43.5 ns 43.5 ns 16141792
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=22.968M/s
FilterBenchmark_versionGreaterEqualFilter/10000/process_time/real_time
129 ns 129 ns 5429183
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=7.74544M/s
FilterBenchmark_versionRangeFilter/10000/process_time/real_time
276 ns 276 ns 2534677
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=3.61734M/s
FilterBenchmark_substringFilter/10000/process_time/real_time
276 ns 276 ns 2534366
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=3.61947M/s
FilterBenchmark_complexFilter/10000/process_time/real_time
28.1 ns 28.1 ns 25516281
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=35.631M/s
```
## After refactor
```
--------------------------------------------------------------------------------------------------------------------------------------
Benchmark
Time CPU Iterations UserCounters...
--------------------------------------------------------------------------------------------------------------------------------------
FilterBenchmark_testStringCFilter/10000/process_time/real_time
15.4 ns 15.4 ns 45535697
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=64.9322M/s
FilterBenchmark_testStringFilter/10000/process_time/real_time
15.4 ns 15.4 ns 45357990
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=64.7906M/s
FilterBenchmark_testLongFilter/10000/process_time/real_time
14.9 ns 14.9 ns 46921381
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=66.9823M/s
FilterBenchmark_testBoolFilter/10000/process_time/real_time
15.4 ns 15.4 ns 45496705
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=64.8934M/s
FilterBenchmark_testVersionFilter/10000/process_time/real_time
19.1 ns 19.1 ns 36715526
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=52.4212M/s
FilterBenchmark_testVersionWithQualifierFilter/10000/process_time/real_time
19.6 ns 19.6 ns 35566450
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=50.8923M/s
FilterBenchmark_testStringGreaterEqualFilter/10000/process_time/real_time
14.8 ns 14.8 ns 47205823
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=67.484M/s
FilterBenchmark_testLongGreaterEqualFilter/10000/process_time/real_time
15.7 ns 15.7 ns 44793529
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=63.8461M/s
FilterBenchmark_testDoubleGreaterEqualFilter/10000/process_time/real_time
16.7 ns 16.7 ns 42395112
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=60.0551M/s
FilterBenchmark_versionGreaterEqualFilter/10000/process_time/real_time
25.6 ns 25.6 ns 27321443
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=39.0526M/s
FilterBenchmark_versionRangeFilter/10000/process_time/real_time
40.9 ns 40.9 ns 17123791
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=24.4385M/s
FilterBenchmark_substringFilter/10000/process_time/real_time
40.8 ns 40.8 ns 17131224
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=24.5143M/s
FilterBenchmark_complexFilter/10000/process_time/real_time
29.3 ns 29.3 ns 23892366
fillEntriesOptimizationBufferPercentage=1
fillStringOptimizationBufferPercentage=0.992188 items_per_second=34.1668M/s
```
# Substring Refactor
The filter substring implementation was incorrect (not according to the
[RFC](https://datatracker.ietf.org/doc/html/rfc4515)) and has been updated.
Additionally, the previous implementation required a malloc during matching;
this need has now been removed, ensuring that a filter match does not require
additional memory and thus cannot fail due to an allocation error.
# Small Refactor of AND and OR
The matching for AND and OR operators has been updated so that `(|)` and
`(&)` will return true.
# Service Registration
Service registration, including component service registration, has been
updated. Now, `service.id` and `service.version` are stored as `long` and
`celix_version_t` types, respectively. Also, some minor "last boy scout rule"
changes have been implemented for error handling in the service registration
flow.
# Error Handling
Most of the string parsing part has been refactored to use `open_memstream`.
Error handling, including error injection testing, is now in place.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
