Greg Landrum
Greg Landrum
Confirmed. Here are some simplified pairs to help with finding the problem (more coming when I work on this): ``` "NC(=[NH2+])c1ccc2cc(C(=O)O)ccc2c1" # fails "NC(=[NH2+])c1ccc2ccc(C(=O)O)cc2c1" # works ```
A test for this is present, commented out, in the changes associated with #8342
Thanks for pointing that out; it was an oversight on my part. I think it makes more sense to just document this and say that atropisomers are only fully supported...
@tadhurst-cdd : when you paste code into GitHub, please surround it with backquotes (if it's inline) or with triple backquotes (if it's multiple lines) so that the format doesn't get...
@AlanKerstjens I like the idea of doing this, and thanks for getting started with it. To answer your questions: using C++17 features is perfectly fine Figuring out how to store...
sorry, clicked the wrong button, I didn't mean to close this.
My two cents on testing this: It's easiest to add C++ tests in the `catch_tests.cpp` file. You can use the tests that are there as examples and just ask if...
A data point on the performance impact of this. I generated 2D coords with and without templates for the first 5K molecules from this set: https://github.com/rdkit/rdkit/blob/master/Regress/Data/znp.50k.smi.gz Timing with master: ```...
Thanks @ZontaNicola , I will take a look at it again in the next couple of days!
> on our testset (~4k molecules) these changes take the number of calls to substructure search from 436 to 103. FWIW I don't see any improvement on my test set....