#8829 closed enhancement (fixed)
Saturation for MWgroups of elliptic curves over number fields.
Reported by:  Robert Bradshaw  Owned by:  John Cremona 

Priority:  major  Milestone:  sage8.1 
Component:  elliptic curves  Keywords:  saturation 
Cc:  Peter Bruin, Kiran Kedlaya  Merged in:  
Authors:  Robert Bradshaw, John Cremona  Reviewers:  David Roe, Kiran Kedlaya, Frédéric Chapoton 
Report Upstream:  N/A  Work issues:  
Branch:  ae6b11c (Commits, GitHub, GitLab)  Commit:  
Dependencies:  Stopgaps: 
Description (last modified by )
Implementation of saturation of points on elliptic curves over number fields. Original patch by Robert Bradshaw in 2010, refactored and made into a git branch by John Cremona in 2015.
I also implemented the simple case of E.gens() for E(K) when E/Q and [K:Q] = 2.
Attachments (1)
Change History (88)
comment:1 Changed 13 years ago by
Status:  new → needs_review 

comment:2 followup: 3 Changed 13 years ago by
I have had a quick look and will go through this in more detail later (after #8828 is completed, probably). I spent a long time on my C++ implementation of this (over QQ but the algorithm is general) so am quite familiar with the details.
Here are two references you should give: [1] S. Siksek "Infinite descent on elliptic curves", Rocky Mountain J of M, Vol 25 No. 4 (1995), 15011538. [2] M. Prickett, "Saturation of MordellWeil groups of elliptic curves over number fields", U of Nottingham PhD thesis (2004), http://etheses.nottingham.ac.uk/52/.
Martin Prickett implemented this in Magma, but the code was very slow and hard to read so it never got incorporated into Magma releases.
Incidentally, it was for this that I implemented group structure for curves over GF(q) in the first place! In my C++ implementation I cache a lot of the information of this group structure so that when you do psaturation for larger and larger p, the structures are already there. A good example is to take one of those curves of very high rank: I think I once successfully psaturated the rank 24 curve at all p < 10^6
(the bound was totally out of reach, something like 10^100
).
Another point which might be useful over number fields: it suffices to use degree one primes to reduce modulo.
Changed 13 years ago by
Attachment:  8829ecnfsat.patch added 

comment:3 Changed 13 years ago by
Replying to cremona:
I have had a quick look and will go through this in more detail later (after #8828 is completed, probably). I spent a long time on my C++ implementation of this (over QQ but the algorithm is general) so am quite familiar with the details.
Here are two references you should give: [1] S. Siksek "Infinite descent on elliptic curves", Rocky Mountain J of M, Vol 25 No. 4 (1995), 15011538. [2] M. Prickett, "Saturation of MordellWeil groups of elliptic curves over number fields", U of Nottingham PhD thesis (2004), http://etheses.nottingham.ac.uk/52/.
Ah, those look like good references to read too :).
Martin Prickett implemented this in Magma, but the code was very slow and hard to read so it never got incorporated into Magma releases.
Incidentally, it was for this that I implemented group structure for curves over GF(q) in the first place! In my C++ implementation I cache a lot of the information of this group structure so that when you do psaturation for larger and larger p, the structures are already there.
The way I do it is consider many p at once, and for each curve over GF(q) I see which primes in my set it could help with, though this won't scale as far. I'm sure there's still lots of room for improvement.
A good example is to take one of those curves of very high rank: I think I once successfully psaturated the rank 24 curve at all p <
10^6
(the bound was totally out of reach, something like10^100
).
That reminds meI was wondering if there's any way to go from min(h(P)) to a bound on the regulator for rank > 1.
Another point which might be useful over number fields: it suffices to use degree one primes to reduce modulo.
Good point. Those get pretty rare for large degree number fields though, right?
comment:4 Changed 13 years ago by
You might also like to look at my C++ code which is in eclib, in src/qcurves. I can point to the right files if it is not clear. In case you wonder, "TLSS" stands for "TateLichtenbaumSamir_Siksek" since I use the TL map when the ptorsion in E(GF(q)) is not cyclic and Samir's original method when it is. Samir only used reduction modulo primes where p exactly divided the order, and in particular for which the reduction had cyclic ppart. But Martin and I discovered that this can fail when there is a pisogeny. Here, fail means in the sense that there can exist points which are not multiples of p in E(QQ) but which map to zero in E(GF(q))/p for all q.
In MP's thesis he proves that this cannot happen if you use all q, or all but a finite number, or all but a finite number of degree 1 primes, .... some of these results we then found had been proved elsewhere (3 or 4 times, independently, within 3 or 4 years!). But it can happen if you leave out the q for which the quotient has noncyclic ppart.
comment:5 Changed 13 years ago by
Authors:  → Robert Bradshaw 

Reviewers:  → John Cremona 
Status:  needs_review → needs_work 
Patch applies fine to 4.4.1 and tests pass.
This functionality is badly needed!
We now have heights for points on curves defined over number_fields but no associated regulator function. I suggest that the function regulator_of_points() be moved up from ell_rational_field to ell_number_field. This tcan then be called instead of the code in lines 424432 [line numbers are from the patched file, not the patch].
Line 439 uses a function self.height_function() which does not exist. This block needs to be replaced by something sensible. If one has a lower bound on the height of nontorsion points, then a bound on the index can be deduced from standard geometry of numbers estimates. To get such a lower bound, see papers of Cremona & Siksek (over Q) and Thongjunthug (over number fields) for an algorithm which would need to be implemented. (Not hard over Q, not much harder for totally real fields, quite a lot worse over fields with a complex embedding). Until this is done, I don't think this saturation function can allow maxprime==0.
In the rank one code: when large primes p (say, over 20!) are being tested then the division_points code will be expensive since it involves constructing the multiplicationbyp map. I would recommend using a reduction strategy here just as in the general case. To check psaturation just find primes q such that #E(Fq) is divisible by p and then see if the reduction of P mod q is a multiple of p. This will almost always prove saturation very quickly. If it fails for several (say 5) q then try to divide P by p; else use more q, and so on. There is one theoretical drawback here: this strategy might fail if there is a rational pisogeny. Over Q, we know which p this might happen for and I would first test for the existence of isogenies of primes degree, and use the division test (as here) for any primes that show up. Over number fields that's harder to deal with, but again we can fall back on the division test to rpove that P cannot be divided by p.
The function list_of_multiples(P,n) duplicated the generic function multiples() which I wrote for just this sort of purpose!
I don't like the loop through all linear combinations for small
primes. Even with p=2 there are curves with 24 independent points out
there and 2^24
divisions is not nice to contemplate. If you want this
short cut, do it based on the size of p^r
.
The main code with reduction etc looks fine to me (but I did not check the logic rigorously).
The gens function for E(K) when E is defined over Q and [K:Q]=2 looks fine. For a more general case we could always try using simon_two_descent (followed by saturation). Trying such an examples led me to:
sage: K.<a> = NumberField(x^22) sage: E = EllipticCurve([a,0]) sage: P = E(0,0) sage: P.has_finite_order() True sage: P.order() 2 sage: P.height() 0 sage: E.saturation([P], verbose=True, max_prime=5) ## infinite loop
This is caused as follows: The height matrix is [0] with det=0 but reg / min(heights) is NaN so reg / min(heights) < 1e6 is False!. This will need fixing. At the very least, I would discard any points of finite order before doing anything else with them. Then min(heights) will never be 0.
Most of the above is easy to deal with. The hard part is computing a suitable max_prime form a lower height bound on points. I suggest that for now you make it compulsory to have a positive max_prime and add a TODO.
comment:6 followup: 7 Changed 13 years ago by
Thank you for all your input. self.height_function
comes from #8828, though as you suggest we could make max_prime mandatory for now (and for rank > 1 once that goes in). That's a good point about large primes in the rank one case. I found the loop through all linear combinations to be much faster in practice for small primes, but the hard coded p == 2
case was left by accident, I meant to cap that on p^r
as I did the others.
I probably won't fix and polish this up before finishing my thesis, but at the latest we should be able to get it done during the workshop at MSRI next month.
comment:7 Changed 13 years ago by
Replying to robertwb:
Thank you for all your input.
self.height_function
comes from #8828, though as you suggest we could make max_prime mandatory for now (and for rank > 1 once that goes in). That's a good point about large primes in the rank one case. I found the loop through all linear combinations to be much faster in practice for small primes, but the hard codedp == 2
case was left by accident, I meant to cap that onp^r
as I did the others.I probably won't fix and polish this up before finishing my thesis, but at the latest we should be able to get it done during the workshop at MSRI next month.
OK  looking forward to it! I'll take a look at #8828 by then as well.
comment:8 Changed 12 years ago by
Since rwb is now busy at Google, and I want this functionality, I am now implementing the changes I suggested above!
comment:11 followup: 12 Changed 10 years ago by
See #12509: until we can fix the height computation, saturation cannot be carried out properly. It's still on my todo list though.
comment:12 Changed 10 years ago by
comment:13 Changed 9 years ago by
Milestone:  sage5.11 → sage5.12 

comment:14 Changed 9 years ago by
Dependencies:  → #8828 

comment:15 Changed 9 years ago by
Summary:  Saturation for curves over number fields. → Saturation for MWgroups of elliptic curves over number fields. 

comment:16 Changed 9 years ago by
Milestone:  sage6.1 → sage6.2 

comment:17 Changed 9 years ago by
Cc:  Peter Bruin added 

comment:18 Changed 9 years ago by
Milestone:  sage6.2 → sage6.3 

comment:19 Changed 8 years ago by
Milestone:  sage6.3 → sage6.4 

comment:20 Changed 7 years ago by
I do not know why this was left drifting, but I really need it myself now so will look at it again, rebase on 6.8 and see what we can do. But I only have one day before a week off, so...
comment:21 Changed 7 years ago by
Authors:  Robert Bradshaw → Robert Bradshaw, John Cremona 

Branch:  → u/cremona/8829 
Commit:  → 0e1e35f624edb087d3fb1ddc21226fec7acfafad 
Description:  modified (diff) 
Keywords:  saturation added 
comment:22 Changed 7 years ago by
Current branch works but more doctests and testing are needed; so not ready for review yet.
I did a lot of rewriting of the main saturation routine, separating off psaturation and also allowing saturation to be done at just one prime. This is a useful special case, since if you take the images of some saturated points under a pisogeny the images may not be psaturated but will still be saturated at all other primes.
The code for computing E(K) when K is quadratic and E is a base change has been completely rewritten and will now work in many more cases (not just when the coefficients of E are in Q).
comment:23 Changed 7 years ago by
Commit:  0e1e35f624edb087d3fb1ddc21226fec7acfafad → a85f40629df89da829e3a38d9e0443576bced01d 

Branch pushed to git repo; I updated commit sha1. New commits:
a85f406  #8829: finished doctests for ec/nf saturation

comment:24 Changed 7 years ago by
The latest commit involves more than adding more doctests to the new functions, since bugs were revealed which led to a rewrite of the sieving code for the two cases where the prank of the reduction is 1 or 2; the former uses discrete log in the reduction, the latter uses Weil pairing and discrte log in the multiplicative group. In the sieve I restrict to primes of degree 1. It is a Theorem (see http://eprints.nottingham.ac.uk/10052/) that this will suffice to prove psaturation, provided that one does use reductions with prank 2 and not just those of prank 1 as originally suggested by Siksek in https://ore.exeter.ac.uk/repository/handle/10871/8323 .
I will mark this as ready for review so the bots get to work on it, and of course humans are welcome to look at the new code, but I will now test it thoroughly on the LMFDB curves and report back.
comment:25 Changed 7 years ago by
Reviewers:  John Cremona 

Status:  needs_work → needs_review 
comment:26 followup: 27 Changed 7 years ago by
Hello,
1) indent correctly the INPUT and OUTPUT fields:
INPUT:  first thing goes one there (not the shift of 2 characters)
2) use the new syntax for raise:
raise MyError("is rich")
Point 1 may be the source of the doc build failure found by the bot:
OSError: [plane_cur] /home/patchbot/sagepatchbot/local/lib/python2.7/sitepackages/sage/schemes/elliptic_curves/ell_number_field.py:docstring of sage.schemes.elliptic_curves.ell_number_field.EllipticCurve_number_field.saturation:9: WARNING: Bullet list ends without a blank line; unexpected unindent.
comment:27 Changed 7 years ago by
Replying to chapoton:
Hello,
1) indent correctly the INPUT and OUTPUT fields:
INPUT:  first thing goes one there (note the shift of 2 characters)2) use the new syntax for raise:
raise MyError("is rich")Point 1 may be the source of the doc build failure found by the bot:
OSError: [plane_cur] /home/patchbot/sagepatchbot/local/lib/python2.7/sitepackages/sage/schemes/elliptic_curves/ell_number_field.py:docstring of sage.schemes.elliptic_curves.ell_number_field.EllipticCurve_number_field.saturation:9: WARNING: Bullet list ends without a blank line; unexpected unindent.
Thanks, I will fix those.
comment:28 Changed 7 years ago by
Commit:  a85f40629df89da829e3a38d9e0443576bced01d → 4f511b89fd199d070bad0d1054efa148d765fdf6 

Branch pushed to git repo; I updated commit sha1. New commits:
4f511b8  fixup doc errors and raise() syntax

comment:29 Changed 7 years ago by
Status:  needs_review → needs_work 

many failing doctests, see bot report
comment:30 Changed 7 years ago by
Apologies, it was a mistake to set this to needs_review prematurely. Next time I do, I will mean it.
comment:31 Changed 7 years ago by
Commit:  4f511b89fd199d070bad0d1054efa148d765fdf6 → 2f20f7373a13e2267bc5ae3d2a7cae93278a76c4 

comment:32 Changed 7 years ago by
Progress report: I am currently running the psaturation (for single primes) on lots of LMFDB curves and all is well so far. This is almost always for very small p (mainly 2 and 3) though, since I am starting with some saturated points on one curve (provided by Magma) and using pisogenies to map to other curves in the isogeny class. Higher degree isogenies are not so common.
I did start to veryify that the points from Magma were fully saturated, but ran into problems computing the saturation index, using (line 3717) the lower bound on the height of all nontorsion points  previously implemented and merged i n6.3 (see #8828). For example, I had a curve where the value of 5 in that line was insufficient *and led to an infinite loop in the call to min()*, while 10 worked fine, but now I have a curve where I have not yet found a value which gives anything. For the record I will give that example here:
K.<phi> = NumberField(x^2x1) # Q(sqrt(5)) E = EllipticCurve([phi + 1, phi + 1, 1, 20*phi  39, 196*phi + 237]) H = E.height_function() H.min(.1,10,verbose=True) # does not appear to terminate
Strictly this is about the code merged in #8828, but it will need fixing here before we can let this (useful!) function out into the world.
comment:33 Changed 7 years ago by
Commit:  2f20f7373a13e2267bc5ae3d2a7cae93278a76c4 → 8686677fa5cdf4359fd6f6b8d8e25925f6893a4c 

Branch pushed to git repo; I updated commit sha1. New commits:
7dbb186  Merge branch 'develop' of trac.sagemath.org:sage into isogs

c66d49f  Merge branch 'develop' of trac.sagemath.org:sage into isogs

ce2472b  Merge branch 'isogs' into sat+isogs

8c80b6f  Merge branch 'saturate' into sat+isogs

8686677  8829: two bug fixes, more tests

comment:34 Changed 7 years ago by
The latest branch I just pushed has some merges in it which were not intended but I hope that will not cause any problems  as well as merging 6.9.beta6 I also merged by branch 'isogs' which has been merged into beta7.
One bug fix addresses the previous comment  after rereading my own 2006 paper I found that the original implementer from #8828 had missed one point (when mu is halved one must increment n_max in order to guarantee termination). A small additional improvement in the same place (the method min_gr() in height.py) now gives a small improvement in the bound, which is why one doctest there has been changed.
The second bug was to do with mutability of lists giving unwanted side effects, and is commented at the point in the source which has changed.
It is likely that users who call the saturation() method will also want to lll_reduce() the output but I have not made that automatic.
I will set this to needs_review once my own full test has completed.
comment:35 Changed 7 years ago by
Status:  needs_work → needs_review 

comment:36 Changed 7 years ago by
Description:  modified (diff) 

Status:  needs_review → needs_work 
After further testing (on many thousands of curves but only psaturating for small p) I saw that it was bad to use discrete_log_lambda() for the dlog in the multiplcative group (in the rarer case where the prank of the reduction is 2 and the Weil pairing is used), both unnecessary and less efficient than simply w.log(zeta). One additional commit coming up...
comment:37 Changed 7 years ago by
Commit:  8686677fa5cdf4359fd6f6b8d8e25925f6893a4c → 99158817c3895207ebae8f3dc06e83f0b7c6a1cb 

Branch pushed to git repo; I updated commit sha1. New commits:
9915881  changed multiplicative dlog to not use discrte_log_lambda

comment:38 Changed 7 years ago by
Status:  needs_work → needs_review 

comment:39 Changed 7 years ago by
Status:  needs_review → needs_work 

There's some bug in the Weil pairing section which I don't have time to fix now, and this has missed 6.9 anyway.
comment:40 Changed 7 years ago by
I did do more work on that but did not get to the bottom of it. Just keeping the ticket alive!
comment:41 Changed 7 years ago by
Commit:  99158817c3895207ebae8f3dc06e83f0b7c6a1cb → f3eb76aaffcd66e55aa48a0ca3fce51f1a4d3a19 

Branch pushed to git repo; I updated commit sha1. New commits:
e481c40  first work on updating #9929 to a working branch (not done yet)

00b6ac6  #8829: ell_height now passes its doctests...

8c44155  #8829: refactored saturation; ell_number_field passes its doctests but more testing and tests needed...

a64e3c9  #8829: finished doctests for ec/nf saturation

573d4b7  fixup doc errors and raise() syntax

7b2f1de  #8829: rewrote projections() so dlogs computed in smaller subgroups

f3eb76a  added check to catch Weil pairing errors during saturation

comment:42 Changed 7 years ago by
Commit:  f3eb76aaffcd66e55aa48a0ca3fce51f1a4d3a19 → a135224f6a7b28c913b4fbe77ade7ed1d15094c3 

comment:44 Changed 7 years ago by
Commit:  a135224f6a7b28c913b4fbe77ade7ed1d15094c3 → 70bb203a14669ff411b2035e33112462c7053707 

Branch pushed to git repo; I updated commit sha1. New commits:
70bb203  fix merge conflicts in ell_number_field

comment:45 Changed 6 years ago by
Branch:  u/cremona/8829 → public/8829 

Commit:  70bb203a14669ff411b2035e33112462c7053707 → 4a7bdc29883e72ceb8e65a8608c6ea3deef4b5fd 
comment:46 Changed 6 years ago by
Commit:  4a7bdc29883e72ceb8e65a8608c6ea3deef4b5fd → 9d66f355dfcba972495f8a459928aeb9488a4baa 

comment:47 Changed 5 years ago by
Commit:  9d66f355dfcba972495f8a459928aeb9488a4baa → a2f50232fc1289a9beec3a488a1cec9907f25f59 

Branch pushed to git repo; I updated commit sha1. New commits:
a2f5023  Merge branch 'develop' into 8829

comment:49 Changed 5 years ago by
Commit:  a2f50232fc1289a9beec3a488a1cec9907f25f59 → 5b40a83c70bee0fc21e1057c2b50475fca5f8a5b 

Branch pushed to git repo; I updated commit sha1. New commits:
5b40a83  #8829: move psaturation code to a new file and fix tests and imports

comment:50 Changed 5 years ago by
Status:  needs_work → needs_review 

After merging into 8.0 I fixed some newly failing doctests (trivial things caused by the usual pari number field changes). I moved the two psaturation functions to a new file saturation.py, and added that to the reference manual. Ready for review. I intend to use this a lot soon over number fields of degree up to 6 for the LMFDB and it would be most helpful to have it merged into 8.1.
comment:51 Changed 5 years ago by
Dependencies:  #8828 

Milestone:  sage6.4 → sage8.1 
comment:52 Changed 5 years ago by
Quick comments :
 dot not use $ but backticks
 every function must be doctested
comment:53 Changed 5 years ago by
!!! New docstrings use backticks, and new functions are doctested. !!! The file ell_number_field is 3800 lines long. Both the files I just worked on have 100% coverage.
comment:54 Changed 5 years ago by
Hello ! some dollars there :
+ For rank 1 subgroups, simply do trial divison up to the maximal + prime divisor. For higher rank subgroups, perform trial divison + on all linear combinations for small primes, and look for + projections $E(K) \rightarrow \oplus E(k) \otimes \FF_p$ which + are either full rank or provide pdivisble linear combinations, + where the $k$ here are residue fields of $K$.
and no doc for
+ def projections(Q, p):
which is indeed an inner function, but quite a complicated one. Maybe just explain its input and output ?
comment:56 Changed 5 years ago by
For the inner function projections(Q, p) there is already a docstring:
Project points onto (E mod Q)(K mod Q) \otimes \F_p. Returns a list of 0, 1 or 2 vectors in \F_p^n
which explains what it does. I'll make sure that all the other inner functions are explained.
comment:57 Changed 5 years ago by
Commit:  5b40a83c70bee0fc21e1057c2b50475fca5f8a5b → 8c715855e54b1b1a12d82364a004f428c9a80097 

comment:58 Changed 5 years ago by
I hope the latest commits do what was wanted for docstrings. I did remove some rather unnecessary oneliner internal functions. In the course of doing this I found at least 2 bugs ;) which is good because the reason this was not finished with 18 months ago was the existence of a bug.
Please review!
comment:59 Changed 5 years ago by
a typo here:
trial divison
and also a missing line break here after r"""
:
+ r""" Given a list of rational points on `E` over `K`, compute the
same here:
"""Return generators for the MordellWeil group modulo torsion, for a
same there:
+ r""" Checks whether the list of points is `p`saturated.
and
+ r""" Full `p`saturation of ``Plist``.
another typo:
divisble
This :
+ if len(EE)==0:
can be replaced by if not EE
another typo:
algirithm
comment:60 Changed 5 years ago by
Commit:  8c715855e54b1b1a12d82364a004f428c9a80097 → 7b3c1886b7dbd481f7248c2271233c73ca483c01 

Branch pushed to git repo; I updated commit sha1. New commits:
7b3c188  #8829 fix docstring typos

comment:63 Changed 5 years ago by
Commit:  7b3c1886b7dbd481f7248c2271233c73ca483c01 → d2421238e6f1f174d90f8f81cbbc0606e79171b3 

Branch pushed to git repo; I updated commit sha1. New commits:
d242123  #8829 correct spelling divison (*4)

comment:64 Changed 5 years ago by
Not now it isn't. I also used "grep r" to catch 3 more (not mine).
One day I might get a review of the code!
comment:65 Changed 5 years ago by
Cc:  Kiran Kedlaya added 

comment:66 Changed 5 years ago by
Some comments on code:
 There's something weird with the indentation at
for Q in K.primes_above(q, degree=1):
inp_saturation
(it looks only indented one space).  There are various points where you don't have spaces around
==
and+=
. If you feel like fixing it, I think that spaces are the Python coding standard.  At various points you add commented out code (either verbose print statements or the definition of
pair_max
). I'm fine with what you have, but I could also see just deleting it. I just wanted to make sure that the decision to include it, commented out, was conscious.
There are also test failures.
Other than these comments, the code looks good to me and I'm happy to give it a positive review once they're addressed.
comment:67 Changed 5 years ago by
Also, the example from comment 5 still seems to go into an infinite loop.
comment:68 Changed 5 years ago by
Status:  needs_review → needs_work 

Thanks to both for the reviews. I'll look into the spacing issues and the example. I really really want to get finished with this!
comment:69 Changed 5 years ago by
I first merged in the current develop (and fixed one small conflict). this required a rebuild (i.e. 'make' not just './sage b' which failed).
I fixed that indentation issue  logically correct but of course nonstandard to have just one space of indentation. I hope the result is OK, that indented block was very long and had a lot of subsidiary indented parts.
I have fixed all (I hope) the == and += spacing.
I changed some commentedout debugging print statements into comments. I like having lots of comments since the logic is quite complicated (and if further bugs are found it may well be me who has to fix them so I might as well be helpful).
I'll test the example from comment 5 once the rebuild has finished.
comment:70 Changed 5 years ago by
Dammit, I always use trac branch names of the form u/cremona/nnnnn where nnnnn is the ticket number, but at some point the branch here became public/8829, so I have just been fixing the wrong version. What a waste of time. Back soon.
comment:71 Changed 5 years ago by
Branch:  public/8829 → u/cremona/8829 

Commit:  d2421238e6f1f174d90f8f81cbbc0606e79171b3 → cd2b023195b23ff4fcf753f67a9acff9c5d87d6b 
Status:  needs_work → needs_review 
comment:73 Changed 5 years ago by
"trial divison" is back, and probably all the other fixes went away.... Sorry for nitpicking, really..
comment:74 Changed 5 years ago by
Status:  needs_review → needs_work 

!@$#!. OK, I will check out commit d242123 by hash to be sure that is what was most recently reviewed, and redo the work I did before. It's the only way to be certain.
comment:75 Changed 5 years ago by
Branch:  u/cremona/8829 → public/8829 

Commit:  cd2b023195b23ff4fcf753f67a9acff9c5d87d6b → 37e22852ecaf466e9d9ce22bc66ffa94f2d7f652 
Status:  needs_work → needs_review 
I hope I got it right this time. The correct branch is public/8829 and I basically redid the edits I had already done this morning on the wrong branch.
Last 10 new commits:
9d66f35  trac 8829 fix for py3

a2f5023  Merge branch 'develop' into 8829

5b40a83  #8829: move psaturation code to a new file and fix tests and imports

8abe3dc  #8829 fix some docstrings and remove some internal functions

847d276  #8829 fix a bug (typo n for p)

8c71585  #8829: put back a oneline internal function

7b3c188  #8829 fix docstring typos

d242123  #8829 correct spelling divison (*4)

63770c3  Merge branch 'develop' into 8829new

37e2285  #8829 after review (2)

comment:76 Changed 5 years ago by
Reviewers:  → David Roe, Kiran Kedlaya, Frédéric Chapoton 

Status:  needs_review → positive_review 
It looks like you've addressed everything. I ran all tests and checked that the loop in comment 5 is no longer a problem.
comment:77 Changed 5 years ago by
Thanks a lot! This ticket was opened in 2010, so I'll be celebrating. Now, if any of you would like to head on over to #22345...
comment:78 Changed 5 years ago by
Very well, and I share your joy, but you introduced a .next(), which is not compatible with python3 (see patchbot plugin report). To be fixed in a later ticket, please.
comment:79 Changed 5 years ago by
That's a pity I thought the next () was rather neat. You can fix it here if you want.
comment:81 Changed 5 years ago by
By the way, John, could you tell LMFDB people about #23671, please ?
comment:82 Changed 5 years ago by
Commit:  37e22852ecaf466e9d9ce22bc66ffa94f2d7f652 → ae6b11c67e43a3a9ffeb3b34d2b46b627967d432 

Status:  positive_review → needs_review 
Branch pushed to git repo; I updated commit sha1 and set ticket back to needs_review. New commits:
ae6b11c  Changed .next() to next() in saturation.py for py3 compatibility

comment:83 Changed 5 years ago by
Replying to chapoton:
no, no, let us wait and do that later
Isn't it just this oneline change? If so, no reason to postpone it...
comment:85 Changed 5 years ago by
Branch:  public/8829 → ae6b11c67e43a3a9ffeb3b34d2b46b627967d432 

Resolution:  → fixed 
Status:  positive_review → closed 
comment:86 Changed 5 years ago by
Commit:  ae6b11c67e43a3a9ffeb3b34d2b46b627967d432 

This is causing doctest failures: #23840.
comment:87 Changed 5 years ago by
I'm pretty certain it will be the usual nonsense of mathematically equivalent outputs. I'll look at #23840 and judge properly.
Some dependance on #8828.