Bug#852780: [sage-packaging] Cantor sage backend
Ximin Luo: > Ximin Luo: >> Antonio Rojas: >>> El lunes, 22 de mayo de 2017 23:59:00 (CEST) Ximin Luo escribió: >>> Thanks for the info. I had a play around, unfortunately --simple-prompt won't be sufficient. > REPLs generally support multiline input (e.g. Python itself) so unless the GSoC student specifically wants Cantor+Sage to work, we'll have to try to figure out how to push ipython to fix their simple-prompt... >>> >>> Thanks, I've added this info to the upstream report >>> https://bugs.kde.org/show_bug.cgi?id=375467 >>> >> >> I think I've managed to fix it in IPython with this patch (2 commits): >> >> https://github.com/ipython/ipython/pull/10606 >> >> Can you confirm? You also need to patch sage/repl/configuration.py as I >> mentioned earlier, as well as patch cantor_sagebackend.so if its version is >> < 16.12. >> > > Unfortunately this broke some Sage doctests. > > To get Cantor to detect Sage properly, I modified IPython's simple-prompt to > not use "In [x]" as the prompt, but to use whatever was configured by the > application (i.e. Sage). > > However, Sage's own doctests assume that IPython would output "In [x]" in > simple-prompt mode. I'm having trouble finding the place in the code where it > does this however, any tips? I confirm that if I patch IPython to hard-code > "In [x]" as the prompt again (whilst also *not* reverting the multi-line fix) > then the sage doctests work again. > Fixed: https://trac.sagemath.org/ticket/23125 -- GPG: ed25519/56034877E1F87C35 GPG: rsa4096/1318EFAC5FBBDBCE https://github.com/infinity0/pubkeys.git
Bug#852780: [sage-packaging] Cantor sage backend
Ximin Luo: > Antonio Rojas: >> El lunes, 22 de mayo de 2017 23:59:00 (CEST) Ximin Luo escribió: >> >>> Thanks for the info. I had a play around, unfortunately --simple-prompt >>> won't be sufficient. >>> REPLs generally support multiline input (e.g. Python itself) so unless the >>> GSoC student specifically wants Cantor+Sage to work, we'll have to try to >>> figure out how to push ipython to fix their simple-prompt... >>> >> >> Thanks, I've added this info to the upstream report >> https://bugs.kde.org/show_bug.cgi?id=375467 >> > > I think I've managed to fix it in IPython with this patch (2 commits): > > https://github.com/ipython/ipython/pull/10606 > > Can you confirm? You also need to patch sage/repl/configuration.py as I > mentioned earlier, as well as patch cantor_sagebackend.so if its version is < > 16.12. > Unfortunately this broke some Sage doctests. To get Cantor to detect Sage properly, I modified IPython's simple-prompt to not use "In [x]" as the prompt, but to use whatever was configured by the application (i.e. Sage). However, Sage's own doctests assume that IPython would output "In [x]" in simple-prompt mode. I'm having trouble finding the place in the code where it does this however, any tips? I confirm that if I patch IPython to hard-code "In [x]" as the prompt again (whilst also *not* reverting the multi-line fix) then the sage doctests work again. X -- GPG: ed25519/56034877E1F87C35 GPG: rsa4096/1318EFAC5FBBDBCE https://github.com/infinity0/pubkeys.git
Bug#852780: [sage-packaging] Cantor sage backend
Ximin Luo: > Antonio Rojas: >> El lunes, 22 de mayo de 2017 23:59:00 (CEST) Ximin Luo escribió: >> >>> Thanks for the info. I had a play around, unfortunately --simple-prompt >>> won't be sufficient. >>> REPLs generally support multiline input (e.g. Python itself) so unless the >>> GSoC student specifically wants Cantor+Sage to work, we'll have to try to >>> figure out how to push ipython to fix their simple-prompt... >>> >> >> Thanks, I've added this info to the upstream report >> https://bugs.kde.org/show_bug.cgi?id=375467 >> > > I think I've managed to fix it in IPython with this patch (2 commits): > > https://github.com/ipython/ipython/pull/10606 > > Can you confirm? You also need to patch sage/repl/configuration.py as I > mentioned earlier, as well as patch cantor_sagebackend.so if its version is < > 16.12. > On my end, it works only half-reliably, but I think the remaining issues are problems with Cantor and the fact it might be trying to use old Sage APIs; not problems with our patched Sage/IPython. >>> integrate(1/(1-x^2), x) - works fine >>> plot(sin(x), (x, 0,2*pi)) - works fine >>> plot(x^3, (x, 0, 5)) - sometimes works fine - sometimes generates an incorrect plot, looks like a scatter plot with random points instead of a cubic curve - sometimes gives "AttributeError: must give both plot endpoints" >>> x = PolynomialRing(RationalField(), 'x').gen() f = (x^3 - 1)^2-(x^2-1)^2 f.factor() - works fine >>> R = PolynomialRing(GF(97),'x') x = R.gen() S = R.quotient(x^3 + 7, 'a') a = S.gen() S - outputs a bunch of unformatted raw latex, even though other examples such as the integrate one above work. in the console where cantor is running, one sees: [..] running: "/usr/bin/dvips" -E -o "/tmp/cantor_tex-V26836.eps" "/tmp/cantor_tex-V26836.dvi" This is dvips(k) 5.996 Copyright 2016 Radical Eye Software (www.radicaleye.com) /usr/bin/dvips: DVI file can't be opened: /tmp/cantor_tex-V26836.dvi: No such file or directory rendered file "/tmp/cantor_tex-V26836.eps" rendered a result to "/tmp/cantor_tex-V26836.eps" settting result to a type 1 result update Entry update result error rendering latex: "something is wrong" [..] X -- GPG: ed25519/56034877E1F87C35 GPG: rsa4096/1318EFAC5FBBDBCE https://github.com/infinity0/pubkeys.git
Bug#852780: [sage-packaging] Cantor sage backend
Antonio Rojas: > El lunes, 22 de mayo de 2017 23:59:00 (CEST) Ximin Luo escribió: > >> Thanks for the info. I had a play around, unfortunately --simple-prompt >> won't be sufficient. >>> >> REPLs generally support multiline input (e.g. Python itself) so unless the >> GSoC student specifically wants Cantor+Sage to work, we'll have to try to >> figure out how to push ipython to fix their simple-prompt... >> > > Thanks, I've added this info to the upstream report > https://bugs.kde.org/show_bug.cgi?id=375467 > I think I've managed to fix it in IPython with this patch (2 commits): https://github.com/ipython/ipython/pull/10606 Can you confirm? You also need to patch sage/repl/configuration.py as I mentioned earlier, as well as patch cantor_sagebackend.so if its version is < 16.12. X -- GPG: ed25519/56034877E1F87C35 GPG: rsa4096/1318EFAC5FBBDBCE https://github.com/infinity0/pubkeys.git
Bug#852780: [sage-packaging] Cantor sage backend
El lunes, 22 de mayo de 2017 23:59:00 (CEST) Ximin Luo escribió: > Thanks for the info. I had a play around, unfortunately --simple-prompt won't > be sufficient. > > > REPLs generally support multiline input (e.g. Python itself) so unless the > GSoC student specifically wants Cantor+Sage to work, we'll have to try to > figure out how to push ipython to fix their simple-prompt... > Thanks, I've added this info to the upstream report https://bugs.kde.org/show_bug.cgi?id=375467
Bug#852780: [sage-packaging] Cantor sage backend
(+CC the debian bug)
Antonio Rojas:
> El lunes, 22 de mayo de 2017 18:56:00 (CEST) Ximin Luo escribió:
>> Has anyone had any success in making this work? In Debian we're suffering
>> from https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=852780 at the moment.
>>
>> I'm not sure how Cantor is interfacing with Sage; and in Debian we do patch
>> away a lot of CLI options relating to Sage-the-distro, similar to what
>> Gentoo does.
>>
>
> Hi,
> I fixed the version detection in 16.12:
>
> https://cgit.kde.org/cantor.git/commit/?id=8d3d07a683ca6758eada1cd8442047401c0fa83d
>
> However, interfacing with Sage is completely broken since the ipython 5.0
> upgrade. Cantor uses KPtyProcess to read Sage's output, and the new ipython
> highlighting confuses it (it can't interpret the ANSI codes). I guess the
> easiest fix would be to add a flag to the sage command that would force
> ipython to run in --simple-prompt mode, and make Cantor run sage with that
> flag.
> Also note that there is an ongoing Google SOC project about rewriting
> Cantor's interface with its backends, so this may be fixed in the process.
>
Thanks for the info. I had a play around, unfortunately --simple-prompt won't
be sufficient.
1. I edited /usr/lib/python2.7/dist-packages/sage/repl/configuration.py to
force _allow_ansi to always return False, which forces simple_prompt=True and
term_title=False.
2. Then I binary-edited the regexp string in
/usr/lib/x86_64-linux-gnu/qt5/plugins/cantor/backends/cantor_sagebackend.so to
instead say '.[a-zA-Z\W].+\s+(\d+)\.(\d+)' - i.e. a hack that effectively
removes the first parentheses so the broken code works "correctly" for 7.6.
Now cantor detects sage's version correctly, but then fails because of
https://github.com/ipython/ipython/issues/9816
found version: ("[.. banner ANSI codes and box drawing characters ..] SageMath
version 7.6", "7", "6")
using the current set of commands
out: "[..] SageMath version 7.6, Release Date: 2017-03-25 [..]\r\nIn [1]: In
[2]: In [3]: In [4]: In [5]: In [6]: In [7]: In [8]: In [9]: TMP_DIR
/home/infinity0/.sage/temp/pdeb1/30547\r\nIn [10]: In [11]: In [12]: File
\"\", line 1\r\ndef
__cantor_enable_typesetting(enable):\r\n
^\r\nSyntaxError: unexpected EOF while parsing\r\n\r\nIn [13]: File
\"\", line 1\r\nif(enable==true):\r\n
^\r\nSyntaxError: unexpected EOF while parsing\r\n\r\nIn [14]: In
[15]: File \"\", line 1\r\nelse:\r\n
^\r\nSyntaxError: invalid syntax\r\n\r\nIn [16]: In [17]: In [17]:
END_OF_INIT\r\nIn [18]: "
You get the same error when doing this manually with sage (with _allow_ansi
hacked to be False as above):
$ sage
[.. banner ..]
In [1]: def __cantor_enable_typesetting(enable):
File "", line 1
def __cantor_enable_typesetting(enable):
^
SyntaxError: unexpected EOF while parsing
REPLs generally support multiline input (e.g. Python itself) so unless the GSoC
student specifically wants Cantor+Sage to work, we'll have to try to figure out
how to push ipython to fix their simple-prompt...
X
--
GPG: ed25519/56034877E1F87C35
GPG: rsa4096/1318EFAC5FBBDBCE
https://github.com/infinity0/pubkeys.git
