Summary
The calculator's dropdown offers 334 stations from /api/stations/, but only 271 have usable pre-calculated percentile data. The remainder fall into two groups:
Even once the 500 is fixed, both groups are selectable options that cannot produce a result.
A station with no data file, selected alongside two that work (#5688):
A station whose data file is entirely null. No error, but no result either:
The 63 with no data file
| Count |
Category |
Examples |
| 20 |
Environment Canada (EC) |
TOFINO AP, VANCOUVER AP, VICTORIA AP |
| 11 |
Other agency (NC) / (MC) / (CRD) |
31N (CRD) 543, 14G (CRD) 1142 |
| 12 |
Test / project stations |
QD 9 TEST, QD1 10M TEST, BLOWDOWN PASS PROJECT |
| 20 |
Other |
TERRACE, KOMIE, BLUEBERRY, COBBLE HILL |
Full list: 1, 2, 26, 44, 49, 69, 83, 88, 122, 142, 157, 186, 203, 205, 229, 395, 434, 435, 480, 543, 896, 915, 967, 1013, 1015, 1016, 1141, 1142, 1143, 1166, 1173, 1851, 3730, 3810, 3851, 3873, 4232, 4270, 4332, 4352, 4393, 4513, 4552, 4653, 4713, 4793, 4973, 5154, 5716, 5796, 5798, 5799, 5996, 5997, 6016, 6076, 6117, 6136, 6156, 6177, 6196, 6216, 6236
Context that may be relevant
app/data/weather_stations.json is still in the repo and still contains exactly 286 entries, matching the 286 data files. The web app also still passes source=unspecified when calling /api/stations/ (PercentileCalculatorPage.tsx:43), which the endpoint no longer reads; StationSourceEnum was removed in #4040. I have not assumed which behaviour is intended.
Approaches
-
Restrict the calculator's dropdown to stations that have usable data. Needs some way for the web app to learn which codes are supported.
-
Regenerate the pre-calculated data to cover more stations. RECENT_YEAR in calculate_percentile_offline.py is currently 2023 and the script reads from an S3 bucket I do not have access to.
Whether these stations should have percentile data at all may be worth a conversation, given that a fair number are Environment Canada, other-agency, or test stations.
It is quite possible I am missing context here, and that some of these are offered deliberately. If so I would be glad to know why. Otherwise, happy to take a run at 1 if that is a direction you would like to go.
Summary
The calculator's dropdown offers 334 stations from
/api/stations/, but only 271 have usable pre-calculated percentile data. The remainder fall into two groups:Even once the 500 is fixed, both groups are selectable options that cannot produce a result.
A station with no data file, selected alongside two that work (#5688):
A station whose data file is entirely null. No error, but no result either:
The 63 with no data file
(EC)(NC)/(MC)/(CRD)Full list: 1, 2, 26, 44, 49, 69, 83, 88, 122, 142, 157, 186, 203, 205, 229, 395, 434, 435, 480, 543, 896, 915, 967, 1013, 1015, 1016, 1141, 1142, 1143, 1166, 1173, 1851, 3730, 3810, 3851, 3873, 4232, 4270, 4332, 4352, 4393, 4513, 4552, 4653, 4713, 4793, 4973, 5154, 5716, 5796, 5798, 5799, 5996, 5997, 6016, 6076, 6117, 6136, 6156, 6177, 6196, 6216, 6236
Context that may be relevant
app/data/weather_stations.jsonis still in the repo and still contains exactly 286 entries, matching the 286 data files. The web app also still passessource=unspecifiedwhen calling/api/stations/(PercentileCalculatorPage.tsx:43), which the endpoint no longer reads;StationSourceEnumwas removed in #4040. I have not assumed which behaviour is intended.Approaches
Restrict the calculator's dropdown to stations that have usable data. Needs some way for the web app to learn which codes are supported.
Regenerate the pre-calculated data to cover more stations.
RECENT_YEARincalculate_percentile_offline.pyis currently 2023 and the script reads from an S3 bucket I do not have access to.Whether these stations should have percentile data at all may be worth a conversation, given that a fair number are Environment Canada, other-agency, or test stations.
It is quite possible I am missing context here, and that some of these are offered deliberately. If so I would be glad to know why. Otherwise, happy to take a run at 1 if that is a direction you would like to go.