Allow floating point numbers for multiplicities #318
Conversation
Codecov Report
Additional details and impacted files |
There was a problem hiding this comment.
In discussions, @bennybp doesn't anticipate any trouble with the database. Thanks for ensuring existing molecule hashes won't change.
It just occurred to me that qcng harnesses could be collecting a non-meaningful float multiplicity that has heretofore gotten cast to int but will henceforth be float unless we do some processing to cast to int for whole numbers. A lot of molecules in the database could be hard to attach new calcs to. Is this a problem that needs a database migration, Ben?
Otherwise, lgtm!
loriab
force-pushed
the
multiplicity
branch
from
ed797b3
to
0106856
Compare
|
@awvwgk, your Note that to create a fractional multiplicity Mol, one has to pass |
|
Why would you have an error for fractional multiplicities when you can pass fractional charges with validation error? |
"... without validation error"? To be clear, mult=3.0 is fine; only values like 3.01 error. From what I recall, the chg/mult logic in molparse wasn't designed for non-int mult or charge. But float charge passes through at least, whereas as-is it errors with float mult -- it would need some changes to accommodate float mult. So how deeply do you want/need fractional quantities incorporated? A small change could probably allow them if fully specified (i.e., no blanks in |
| if isinstance(molecular_multiplicity, float) or any(isinstance(m, float) for m in fragment_multiplicities): | ||
| raise TypeError( | ||
| f"validate_and_fill_chgmult() cannot handle fractional multiplicity. m: {molecular_multiplicity}, fm: {fragment_multiplicities}" | ||
| ) |
There was a problem hiding this comment.
If this is not blocking the use of multiplicities like 1.5 I am fine with merging this change. I don't see an equivalent check for charges therefore I am a bit concerned whether this change only in principle allows fractional multiplicities.
|
Lori @loriab, can we merge this PR or is anything blocking it? |
|
What is the status of this PR, is anything missing? |
|
Sorry for the delay. So long as what you need is unvalidated pass-through of non-int chg and mult, I think all this PR needs is an extra message for charge like the new warning one for mult in the case of non-integers. I'd better add some non-int charge tests, too. It's true that float charge has been there from the beginning but possibly never exercised. |
|
I think that works for now as a solution |
|
Ok, loriab@342d7cc is my current proposal for allowing float multiplicity. It's sitting atop #343 for now. Investigation showed that the validation routine does handle float charges and checking internal consistency between total and fragment charges. It doesn't try to assign fractional multiplicities to go along with fractional charges, and this is probably edge-case enough in QC to not demand those checks. Accordingly, the validation routine has been made float-tolerant for multiplicities, so validate=T/F allowed. It doesn't try to do any fancy guessing, and you may find cases where its handling should be loosened, tightened, or better coordinated with charge in future. |
Closes #317
Description
Changelog description
Status