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:
https://github.com/tmdigert/simple-bleve-search
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:
As compared to running the same test in 2.4.3:
Using defined numbers of entries in the index in 2.5.2 (Batch size of 100):
As compared to running the same test in 2.4.3:
With varying test polygons that return smaller numbers of hits in 2.5.2 (Batch size of 100):
As compared to running the same test in 2.4.3:
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:
https://github.com/tmdigert/simple-bleve-search