-
Notifications
You must be signed in to change notification settings - Fork 697
Description
This issue essentially extends issue #2114, since that issue was closed it appears that a number of commits regarding GeoShape functionality have been introduced. However, I'm now getting considerably longer query times than what was seen in #2114 on the current release of Bleve (2.5.2). I was able to produce times similar to that previous issue on my end when using 2.4.3. I re-created that earlier test using the code provided and with a .brotli dump from the same endpoint.
Searching for the same polygon with different batch sizes resulted in the following times in 2.5.2:
| Batch | >10000 | >5000 | >1000 | >500 | >100 | >50 |
|---|---|---|---|---|---|---|
| ms | 701 | 650 | 756 | 678 | 738 | 676 |
As compared to running the same test in 2.4.3:
| Batch | >10000 | >5000 | >1000 | >500 | >100 | >50 |
|---|---|---|---|---|---|---|
| ms | 265 | 283 | 279 | 284 | 248 | 247 |
Using defined numbers of entries in the index in 2.5.2 (Batch size of 100):
| Entries | 100,000 | 200,000 | 300,000 | 400,000 | 534,040 |
|---|---|---|---|---|---|
| ms | 157 | 219 | 433 | 639 | 724 |
As compared to running the same test in 2.4.3:
| Entries | 100,000 | 200,000 | 300,000 | 400,000 | 534,040 |
|---|---|---|---|---|---|
| ms | 53 | 101 | 135 | 177 | 239 |
With varying test polygons that return smaller numbers of hits in 2.5.2 (Batch size of 100):
| Polygon | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| ms | 744 | 760 | 769 | 749 | 710 |
| hits | 51080 | 35820 | 12483 | 367 | 6 |
As compared to running the same test in 2.4.3:
| Polygon | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| ms | 240 | 245 | 264 | 245 | 225 |
| hits | 51080 | 35820 | 12483 | 367 | 6 |
Please let me know if I'm doing something incorrect or if there's a better approach. Here's a link to my test code: