These are chat archives for spyder-ide/public

25th
Apr 2019
Luca
@lucamerio89_gitlab
Apr 25 08:37
Hi Everybody (this is my first message here). I noticed that the code analysis returns a red error with the code def my_sum(a: 'first addend', b: 'second addend'): return a+b. Apparently it doesn't like the two functions annotations there. However the code is perfectly legit and runs. The same "false error" is returned when copy-pasting the very first example in PEP-3107. Is there a good reason or is this an error in the code analysis?
Mark B. Campanelli
@campanelli-resonon
Apr 25 14:56
@CAM-Gerlach Thanks for the info on the situ. My PySide2 preference is because I use PySide2 in a smaller project for a new product. I like it and the license is a bit nicer than PyQt. The app discussed here still uses WX, but at this point I see no reason that I cannot run spyder embedded in its own process alongside it. spyder has been the most promising Python3 replacement for our embedded Python Toolkit (PTK) IDE that I have found so far. It's too bad the pip installation of spyder for Python2 is so heavyweight due to PyQt5, but I have verified that it is much more straightforward in a Python 3.6.8 venv. (FYI I tend to prefer conda, but I'm an exception in our org. Actually, one of my biggest issues with conda is that matplotlib seems to demand PyQt, and it's at a somewhat stale version at that. That out-of-the-box dependency bloat is no good for automated production CI workflows for packaged software delivery.) I am now wondering about the license for PyQt5 if I embed spyder in our closed source app. I'm not totally sure how that applies if I'm using a derivative software. (It appears that spyder itself is under MIT.)
Gonzalo Peña-Castellanos
@goanpeca
Apr 25 15:16
Now that you mention that topic, I knew there were plans at some point to split the matplotlib packages to avoid having it depend on qt directly. But that would rely on a concept that does not exist in conda (optional dependencies)
This has been discussed in the past
But I am afraid there has not been much progress so far
Conda is a bit underresourced I bleieve and they (Anaconda, inc) are trying to keep up with issues about speed etc.
Also why you see some alternatives like https://medium.com/@wolfv/making-conda-fast-again-4da4debfb3b7
Which might make it uptream
CAM Gerlach
@CAM-Gerlach
Apr 25 16:21

@campanelli-resonon Similarly, having Spyder depend optionally on either PyQt5 or PySide2 would require that as well. Before, we just told users to install one separately, but naturally many users would not RTFM and complain that their Spyder was broken since they hadn't installed a Qt binding.

@goanpeca Wow, that is blazing fast! The non-solver stuff is still way slower on Windows than on my Linux VM running in ye olde Virtualbox on top of the same machine (I've noticed the same thing with git, find and other commands), but on Linux the only thing I really spend time waiting for is the solver.

Mark B. Campanelli
@campanelli-resonon
Apr 25 16:44
@goanpeca Thanks for sharing that Medium article, and I'm glad that I am apparently not missing something w.r.t. conda's strong coupling of matplotlib and qt. That said, I understand that conda's approach w.r.t. matplotlib is probably great for is main user base, and conda's virtual environment setup for building extension modules on Windows blows away virtualenv/venv IMHO.