Peter Johnson

Results 1014 comments of Peter Johnson

Thanks for the bug report. The suburb of Ballina has an incorrect synonym of '[West] Ballina' which confuses it with a nearby suburb of the same name. I have opened...

this was resolved in https://github.com/whosonfirst-data/whosonfirst-data-admin-au/pull/49 it will take a week or so to reflect in the production service from [geocode.earth](https://geocode.earth/). @stinnock are you using the Geocode Earth service or a...

I upgraded our spec file recently, this was the diff:

The setting is already [configurable here](https://github.com/pelias/spatial/blob/656fa8d5100ad1c071632f732840b74f4c345940/module/shard/TriggerGeometryInsert.js#L8) and [overridable via the CLI](https://github.com/pelias/spatial/blob/656fa8d5100ad1c071632f732840b74f4c345940/bin/cmd/import.js#L58C19-L58C48) with a command such as `... import --tweak_module_shard_complexity=2000 zcta`

I just used zcta because it's small and builds in like a minute on my laptop, my latest PR has some new notes on how to generate the data, and...

Here's a script which can be used to generate a bunch of variations of shard count complexity: ```bash sqlite3 /data/wof/whosonfirst-data-admin-nz-latest.db 'SELECT json_extract(body, "$") FROM geojson' | tee \ >(node bin/spatial.js...

I did some basic `k6` load testing of these file against the more complex `/query/pip/_view/pelias` endpoint with the following results: ```bash complexity 200 avg=6.04ms min=1.09ms med=5.28ms max=51.65ms p(90)=9.96ms p(95)=11.62ms avg=6.19ms...

As expected, we see the filesize reduce as the shard complexity rises, likewise we see an increase in latency with a cliff at the higher complexities: Looks like maybe a...

Also worth noting that running these benchmarks on my laptop it's going to be tricky getting a good reading, I turned off `info` level logging in `pelias.json` to avoid seeing...

Some more interesting stats, looking at the average amount of shards produced per-geometry with the different shard complexity settings: It shows that even the setting `100000` is generating a bunch...