Depiction of large structures makes Python crash on Windows

[OBrink]

Hey there!

We have run into a problem with the depiction functionalities in PIKAChU. When depicting large molecules, Python simply crashes. We don't get an exception that can be caught or something like that. If I execute the code below in my Python shell, the Python shell simply closes.

Minimal example for the reproduction of the problem:

from pikachu.general import read_smiles
from pikachu.drawing import drawing

smiles = 'CC(=O)N[C@H]1[C@H]([C@H](O)[C@H](O)CO)O[C@@](O[C@H](CO)[C@@H](O)[C@@H]2O[C@@](O[C@H](CO)[C@@H](O)[C@@H]3O[C@@](O[C@H](CO)[C@@H](O)[C@@H]4O[C@@](O[C@H](CO)[C@@H](O)[C@@H]5O[C@@](O[C@H](CO)[C@@H](O)[C@@H]6O[C@@](O[C@H](CO)[C@@H](O)[C@@H]7O[C@@](O[C@H](CO)[C@@H](O)[C@@H]8O[C@@](O[C@H](CO)[C@@H](O)[C@@H]9O[C@@](O[C@H](CO)[C@@H](O)[C@@H]%10O[C@@](O[C@H](CO)[C@@H](O)[C@@H]%11O[C@@](O[C@H](CO)[C@@H](O)[C@@H]%12O[C@@](O[C@H](CO)[C@@H](O)[C@@H]%13O[C@@](O[C@H](CO)[C@@H](O)[C@@H]%14O[C@@](O)(C(=O)O)C[C@H](O)[C@H]%14NC(C)=O)(C(=O)O)C[C@H](O)[C@H]%13NC(C)=O)(C(=O)O)C[C@H](O)[C@H]%12NC(C)=O)(C(=O)O)C[C@H](O)[C@H]%11NC(C)=O)(C(=O)O)C[C@H](O)[C@H]%10NC(C)=O)(C(=O)O)C[C@H](O)[C@H]9NC(C)=O)(C(=O)O)C[C@H](O)[C@H]8NC(C)=O)(C(=O)O)C[C@H](O)[C@H]7NC(C)=O)(C(=O)O)C[C@H](O)[C@H]6NC(C)=O)(C(=O)O)C[C@H](O)[C@H]5NC(C)=O)(C(=O)O)C[C@H](O)[C@H]4NC(C)=O)(C(=O)O)C[C@H](O)[C@H]3NC(C)=O)(C(=O)O)C[C@H](O)[C@H]2NC(C)=O)(C(=O)O)C[C@@H]1O'

mol = read_smiles(smiles)
drawer = drawing.Drawer(mol)

Expected behaviour:

• Instantiation of a Drawer object.

Observed behaviour:

• Python crashes.

We will build a workaround and simply not use PIKAChU for large structures in our application, but I think it may make sense to look into this.

Have a nice day!
Otto

[BTheDragonMaster]

Hi Otto,

Thank you for reporting this!

On my personal laptop (Mac with M1 chip), PIKAChU (1.0.10) does draw the structure (although admittedly it takes around half a minute):

from pikachu.general import svg_from_smiles
svg_from_smiles(r"CC(=O)N[C@H]1[C@H]([C@H](O)[C@H](O)CO)O[C@@](O[C@H](CO)[C@@H](O)[C@@H]2O[C@@](O[C@H](CO)[C@@H](O)[C@@H]3O[C@@](O[C@H](CO)[C@@H](O)[C@@H]4O[C@@](O[C@H](CO)[C@@H](O)[C@@H]5O[C@@](O[C@H](CO)[C@@H](O)[C@@H]6O[C@@](O[C@H](CO)[C@@H](O)[C@@H]7O[C@@](O[C@H](CO)[C@@H](O)[C@@H]8O[C@@](O[C@H](CO)[C@@H](O)[C@@H]9O[C@@](O[C@H](CO)[C@@H](O)[C@@H]%10O[C@@](O[C@H](CO)[C@@H](O)[C@@H]%11O[C@@](O[C@H](CO)[C@@H](O)[C@@H]%12O[C@@](O[C@H](CO)[C@@H](O)[C@@H]%13O[C@@](O[C@H](CO)[C@@H](O)[C@@H]%14O[C@@](O)(C(=O)O)C[C@H](O)[C@H]%14NC(C)=O)(C(=O)O)C[C@H](O)[C@H]%13NC(C)=O)(C(=O)O)C[C@H](O)[C@H]%12NC(C)=O)(C(=O)O)C[C@H](O)[C@H]%11NC(C)=O)(C(=O)O)C[C@H](O)[C@H]%10NC(C)=O)(C(=O)O)C[C@H](O)[C@H]9NC(C)=O)(C(=O)O)C[C@H](O)[C@H]8NC(C)=O)(C(=O)O)C[C@H](O)[C@H]7NC(C)=O)(C(=O)O)C[C@H](O)[C@H]6NC(C)=O)(C(=O)O)C[C@H](O)[C@H]5NC(C)=O)(C(=O)O)C[C@H](O)[C@H]4NC(C)=O)(C(=O)O)C[C@H](O)[C@H]3NC(C)=O)(C(=O)O)C[C@H](O)[C@H]2NC(C)=O)(C(=O)O)C[C@@H]1O", 'large_mol.svg')

I've also run the snippet of code you sent, and while it takes some time to run, it does return the drawer object which behaves normally on my end. I appreciate that the performance of the M1 chip is not really comparable with a lot of processors out there, so I will test PIKAChU's performance on a range of other processors to see what happens.

I will have some time to work on PIKAChU later in December, and will look at this issue more deeply then. Again, thank you for pointing this out!

Kind regards

[OBrink]

Thanks for the quick reply. I am running this on my laptop with an Intel(R) Core(TM) i7-10750H CPU @ 2.60GHz 2.59 GHz processor and 32 GB RAM. I doubt that this is hardware-related, and I have found something that supports this claim:

When I run the code snippet above under Windows 11, everything crashes. @Mar-Gol observed the same behaviour under similar conditions. When I run it in my Ubuntu subsystem in the WSL shell, everything works fine. I just ran it remotely on an Ubuntu server, and it also works fine. This seems to be a Windows-specific problem, but I have absolutely no idea what the root of it is.

[BTheDragonMaster]

Thank you for testing this - I will have a look at what might cause the different behaviours across different operating systems.

[DocMinus]

Hi - I can chime in that it works under Ubuntu on a Xeon multicore system.
On mine,both examples take about 60-65 secs.
Tested also on Python 3.11 which is supposed to be faster, and yes, there it takes about 40ish secs.

Finally did a quick test on a win10 machine, and yes, it doesn't work there. Doesn't crash my terminal, but the script ends within a few seconds.