Added new class method to HLSCatalog #9
Conversation
| tiles = df.apply(lambda row: lookup.get_point_hls_tile_id(row.lat, row.lon), axis=1) | ||
| return self.from_tiles(tiles, years, bands, tile_lookup) |
There was a problem hiding this comment.
One of the (unstated) goals of this function was to retain metadata provided in the df parameter (such as training id). This isn't preserved here because from_tiles operates only on a series of tiles.
Also there are no years passed into from_point_pandas because it's required that the year is a column in the input df.
One way we can use your approach is:
- get the series of tiles for the points and add it as a column to the input
df. - sub-select two columns from
df: year and tile - remove duplicates the sub-df
- adapt
from_tilesto accept a dataframe with tile+year rather than accepting tiles and years separately - Call
from_tileson the sub-df in (3) - Join the resulting dataset from
from_tileswith the point dataframe on year/tile
This saves us duplicate computation for points with the same tile/year and means we can share code with from_bbox
There was a problem hiding this comment.
Thanks, I see now this didn't work. I really just wanted a from_tiles fxn to test with and wanted to reduce duplication, but....
Are you in a situation where you need different years of data for different points?
There was a problem hiding this comment.
Yeah one dataset has samples from different years
|
Closing for now, non-essential |
Sorry, should have put this on another branch?