Tests for lcalc files

[davidfarmer]

If we are going to have links to Lcalc files, then we need to have tests for them.

[JohnCremona]

What sort of tests? Adding tests to make sure that the links work, that's easy. Testing that the code then runs is much harder and might be impossible. We try (I hope) to make sure that the Sage / Magma / GP code does work, but I do not know a way to automate that and occasionally there is a problem.
I'm sure that there is someone out there who could write a script which would download some of these code files and run the code and compare the output with a precomputed standard, which could be run at least locally. Getting Travis to do this is a rather harder job (and impossible for Magma, not being free or easy to install).

[edgarcosta]

I think we should serve the Lcalc files with a disclaimer along the lines:

Many of the properties of the L-functions are pre-computed and require many more Dirichlet coefficients than the ones stored on our database, hence the ones provided in this Lcalc file.

One could also add a line saying that these are just meant as mock up examples.

[davidfarmer]

Closing, because we don't want any links to lcalc: all the L-function values should be served
from the database.

We could open another issue to eliminate lcalc, but that probably is not necessary because
it should be automatic once all the L-function data is precomputed.

[davidfarmer]

Reopening because there should be some discussion about whether lcalc files or an alternative
should be available as a way to download that data.

[AndrewVSutherland]

Perhaps these should be viewed in the same category as the code snippets we have on many pages, which provide a way for a user to reconstruct an object or some of its attributes in an external software package, but these code snippets do not necessarily represent the method used to compute the data that appears on the page. For example, on an elliptic curve page you can click the Magma button to get a code snippet to compute generators for the Mordell-Weil group in Magma, but that is not how the generators stored in the LMFDB were computed, and if if your this code in Magma you will not necessarily get the same generators that appear on the page. But I think the code snippet is still useful.

[edgarcosta]

I also think these should be viewed in the same category as the code snippets.

…

On 23 May 2018 at 09:58, Andrew Sutherland ***@***.***> wrote: Perhaps these should be viewed in the same category as the code snippets we have on many pages, which provide a way for a user to reconstruct an object or some of its attributes in an external software package, but these code snippets do not necessarily represent the method used to compute the data that appears on the page. For example, on an elliptic curve page you can click the Magma button to get a code snippet to compute generators for the Mordell-Weil group in Magma, but that is not how the generators stored in the LMFDB were computed, and if if your this code in Magma you will not necessarily get the same generators that appear on the page. But I think the code snippet is still useful. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1257 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AATtBlgDSo5abGwH0-n_WOeeIKb0B3krks5t1WrpgaJpZM4IabMH> .

[edgarcosta]

Should we add a disclaimer to the header of the files?

[AndrewVSutherland]

At the LMFDB editorial meeting in August we agreed to remove the L-calc links, but this won't happen until after the postgres conversion is complete and the new modular forms code with precomputed L-functions is live. But I'm happy to leave this issue open until the situation is resolved. There is nothing wrong with adding a disclaimer in the mean time if someone wants to do it.

[edgarcosta]

Great. That is what I was hoping for. I will remove them together with the other changes associated to classical modular forms.

…

On Sat, Oct 6, 2018, 12:10 Andrew Sutherland ***@***.***> wrote: At the LMFDB editorial meeting in August we agreed to remove the L-calc links, but this won't happen until after the postgres conversion is complete and the new modular forms code with precomputed L-functions is live. But I'm happy to leave this issue open until the situation is resolved. There is certainly fine with adding a disclaimer in the mean time if someone wants to do it. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1257 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AATtBoC2iAQh6we0jX7klJk4oxFLK7Duks5uiNXggaJpZM4IabMH> .

[AndrewVSutherland]

With #2717 merged the postgres version is now running on www.lmfdb.org and L-calc links no longer appear.