Add a record for qchem geometry optimization

[mattwelborn]

I don't have a license...

[dgasmith]

Are you looking for the output file of a geometry optimization or two? Any particular molecule?

[mattwelborn]

Something small like water with a short trajectory to make it easy to write the reference qcschema, please.

…

On Mar 7, 2020, at 17:17, Daniel Smith ***@***.***> wrote:  Are you looking for the output file of a geometry optimization or two? Any particular molecule? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[dgasmith]

For parsing optimizations from pure log files you might want to look in cclib. You start to need state machines to parse this kind of output well.

output.txt

[mattwelborn]

Thanks. Out of curiosity, why don't we just use cclib as the base parser for qcng and supplement when additional data are available? (e.g. read the qchem output log with cclib, translate to qcsk, then read the scratch folder with qcng)

[dgasmith]

Generally cclib doesn't parse enough to be valid in schema land so we need to reparse the majority of the quantities regardless. The primary exception to this is their parsing of composite quantities like geometry optimizations where most codes do not have binary representations thereof.

[mattwelborn]

Okay. In that case, it doesn't seem worth introducing a cclib dependency. In the past, I wrote a state machine parser for qchem output, but I think we can skate by with a very simple state machine + what we already have. Basically, use a regex to split the geometry optimization frames apart and parse each one as an AtomicResult. Any objections?

[dgasmith]

Not on my end, Q-Chem has a nice regular structure to it. Some other programs use different print outs for individual gradients which makes things difficult.