Closed low-decarie closed 1 year ago
SCISAT refreshes within 10-12 seconds now on the QA server. This is not perfect, but a great improvement. The source code contains TODO statements that identify the slowest parts.
numpy.nan
is slowTo improve the performance, I added a static variable - data_reader.page_df
- in data_reader()
. It makes sure the function will not be called 3 times with the same parameters. In addition, I speed up this part:
Before:
for i in range (nbDays): date.append(dt.datetime(years[i],months[i],days[i]))
Now:
date = np.array([dt.datetime(years[i],months[i],days[i]) for i in range (nbDays)])
Bassically speaking, two methods are still slow:
data_reader()
(because of numpy.nan)generate_geo_map()
(because of the bin mapping)The latest version of SCISAT was deployed on the production server on May 24th, 2023. This new version is much faster than the previous one. The Graphical User Interface (GUI) hasn’t changed. https://donnees-data.asc-csa.gc.ca:8888/scisat/
Same as https://github.com/asc-csa/AlouetteApp/issues/136