You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
import astroplan should not access the internet #15
@eteq@bmorris3 I think the fact that astroplan downloads the IERS table in import is a temp solution anyways. Is there any way to set this up better for now? (if no, no worries, we'll just uncomment that line in Gammapy).
I think the fact that astroplan downloads the IERS table in import is a temp solution anyways. Is there any way to set this up better for now?
That's correct, it's a temporary solution to force astropy to always use the most up-to-date IERS table. The future solution will only download the newer table when recent dates are queried and the most recently downloaded IERS A table is out of date.
Since it might take some time to sort out the astropy.time adjustments necessary to accommodate that goal, and since this temporary solution simulates what we want the corrected version of astropy to do, I've moved ahead on developing astroplan (See #14 ).
I would like to ask @eteq again if it would be reasonable to include the IERS Bulletin A in astroplan. From the Bulletin A available versions page, it looks like the table is updated once a week but I can't imagine that the precision gained from such frequent updates are truly necessary for the average user.
We could include a recent version of the bulletin with the install and raise a warning when the cached table is more than a week old and the internet cannot be accessed to grab the latest bulletin. For users who can't access the internet, the best you can do is use the most recent copy from the last time you were connected to the internet. And if you're never connected to the internet apart from the time when you downloaded astroplan, the best you can do is get a table contemporary with your version of the astroplan repo.
Of course, the above suggestion would work best if there were a web hook or similar service that updated the table file every time we pushed/merged an update to astroplan. @adrn / @cdeil / @eteq Does a service like that exist, or am I dreaming in gitopia?
That's a 4 MB file?
If so (or smaller), I think it's OK to bundle a version if it resolves this issue.
Is it needed for a lot of functionality?
If no, another solution (which is probably less controversial) is to make it optional, i.e. make import astroplan and python setup.py test run even if the file is not available (skipping the parts that need that file).
So this issue with downloading the IERS table on first import was temporary and must have been due to availability issues of the download site or travis-ci.
I'll re-start the builds on the open PRs that had this fail now ...
Activity
bmorris3 commentedon Jun 28, 2015
That's correct, it's a temporary solution to force astropy to always use the most up-to-date IERS table. The future solution will only download the newer table when recent dates are queried and the most recently downloaded IERS A table is out of date.
Since it might take some time to sort out the
astropy.timeadjustments necessary to accommodate that goal, and since this temporary solution simulates what we want the corrected version of astropy to do, I've moved ahead on developing astroplan (See #14 ).cdeil commentedon Aug 21, 2015
As pointed out in #42 (comment), this issue currently makes all tests on travis-ci fail and prevent progress on other PRs.
I think it's really important to
import astroplanshould always work without internet access@bmorris3 or @eteq – Do you have a proposal or PR for a solution to this issue?
bmorris3 commentedon Aug 21, 2015
I would like to ask @eteq again if it would be reasonable to include the IERS Bulletin A in astroplan. From the Bulletin A available versions page, it looks like the table is updated once a week but I can't imagine that the precision gained from such frequent updates are truly necessary for the average user.
We could include a recent version of the bulletin with the install and raise a warning when the cached table is more than a week old and the internet cannot be accessed to grab the latest bulletin. For users who can't access the internet, the best you can do is use the most recent copy from the last time you were connected to the internet. And if you're never connected to the internet apart from the time when you downloaded astroplan, the best you can do is get a table contemporary with your version of the astroplan repo.
bmorris3 commentedon Aug 21, 2015
Of course, the above suggestion would work best if there were a web hook or similar service that updated the table file every time we pushed/merged an update to astroplan. @adrn / @cdeil / @eteq Does a service like that exist, or am I dreaming in gitopia?
cdeil commentedon Aug 21, 2015
That's a 4 MB file?
If so (or smaller), I think it's OK to bundle a version if it resolves this issue.
Is it needed for a lot of functionality?
If no, another solution (which is probably less controversial) is to make it optional, i.e. make
import astroplanandpython setup.py testrun even if the file is not available (skipping the parts that need that file).cdeil commentedon Aug 24, 2015
It looks like travis-ci builds for astroplan are passing again (see #68 and https://travis-ci.org/astroplanners/astroplan/builds/77018060) without any changes on our side.
So this issue with downloading the IERS table on first import was temporary and must have been due to availability issues of the download site or travis-ci.
I'll re-start the builds on the open PRs that had this fail now ...
😅
cdeil commentedon Aug 28, 2015
This issue has been resolved in #71 ... closing now.