-
Notifications
You must be signed in to change notification settings - Fork 346
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
grdinfo on tiled files may be overkill #3767
Comments
I agree that assembling tiles to get min max of entire dataset is silly. It should only happen with a non-global -R. So we would have to add zmin/zmaz to the gmt_data_server.txt file and that means 6.2 I think . |
Another thought I had about tiles is this: If the user downloads all the tiles for a resolution, let grdblend make the single earth_relief_xxy.grd file and then we use that file instead (if it exists). This would eliminate lots of repeated grdblend calls, such as happens every time you run a script that makes a map. I think once the large file is there, it is simpler to just pass w/e/s/n in a read like we have always done. Should we pursue that scheme? |
Yes, it's much better to have a single file that playing with the tiles all the time. For example
will launch a grdblend job twice, whilst with a single file it's just netcdf reading sub-regions. |
We can update the gmt_server_data.txt with two more columns for zmin and zmax. However, this will break backwards compatibility, so I think we will need to do this (assume we may make other changes in the future);
I think that is better since we may not make more changes and then we dont have to add copies of files with 6.2. 6.3 etc if no changes. Other solutions? |
As for zmin and zmax, do we really need it? It's only useful for |
Good point, that is true, and while @joa-quim may have forgotten the range of Earth's relief, most grade school kids know it goes from -11km to 8 km, ish. So perhaps it is not that helpful information to encode, given the trouble it imposes. Perhaps the more helpful upgrade would be to reassemble the full global grids when the user gets all the tiles of a kind. |
Yes, I think it's a good idea, but it means double the disk usage, and for earth relief 01s, it's a huge file (~40GB) (but will anyone plot a global map using 01s data?). |
Well, right now there is no global tile set so even if you got all the SRTM1 tiles you still dont have 360x180 of them. So 1s and 3s will not change. It is more the intermediate such as 15s 30s and possibly 1m that are still too large to ship as a single item, but at least the 1m I can easily see be quickly downloaded via tiles and reach full coverage. My thought (have not thought very long on this yet) would be that when gmt_assemble_grid is called on tiles, it could check if we have full coverage, and if we do then we launch a job to build the single global file. Not sure of details though: SHould we then delete the tiles if the user has earth_relief_30s.grd? Doing the blend takes some time for these large files, is the user willing to just wait for that? Or do we simply explain all this better and offer this via a gmt get option of modifier. Perhaps gmt get, which gets all the tiles of the selected resolutions, should simply build the single grid instead? |
And also how to update the single global file if the data has a new release. |
First let me say that I found this because some julia function that automatically fishes the grid extents did a Tiles are very convenient to work on small and very variable regions but are not so good to work with, specially if their number starts to grow. I think the best would be, if for any reason, including, accumulated usage some say 75% percent of the tiles end up in local machine, then better download them all and recreate the monolithic file and DELETE the tiles. When data has a new release, for the monolithic, files should NOT be updated automatically but a warning about the new version should be issued, together with the |
I will wait a bit on this to see if the NASA proposal gets funded which is were we proposed the tile mechanism that we now actually have started to implement. |
For example, this one already take a lot of time and uses an extra 1 GB of mem
The problem is obviously because it has to assemble all the tiles to compute the info. A possible solution would be to store this info together with the files and add another cleverness to the reader to consult that file first if no -R is provided.
The text was updated successfully, but these errors were encountered: