Passing PMCID to PMID citation parser

[agitter]

I wanted to use Manubot to cite PMCID PMC6063279 and instead received a citation for PMID 6063279. I had accidentally written

manubot cite pmid:PMC6063279

Apparently the Entrez eutils takes the PMCID, strips PMC, and then returns the XML for that numeric PMID. This seems like an error on the part of eutils. Do we want to try to prevent this type of user error? Or should we instead leave the Manubot behavior but notify the eutils maintainers?

[dhimmel]

Ooh that's an interesting behavior to strip PMC.

Do we want to try to prevent this type of user error?

Yes, we could check the pubmed ID for being all numeric.

This seems like an error on the part of eutils

I have found it's difficult to get in touch with the right developers at NCBI. For example, see https://github.com/ncbi/citation-exporter/issues/3: I created a GitHub issue, but I was still told to go through Helpdesk tickets, where I never received a response that actually acknowledged the feature request.

[dhimmel]

Apparently the Entrez eutils takes the PMCID, strips PMC

Also sounds like whoever engineered this considered the character stripping a feature... so it's likely not considered a bug, and is perhaps utilized by other applications (;

[agitter]

Yes, we could check the pubmed ID for being all numeric.

What would be more user-friendly?

• Throw an error if someone uses pmid:PMC6063279
• Check whether the non-numeric PMID starts with PMC and redirect the citation to get_pmc_citeproc

[dhimmel]

We could warn and redirect, but I'm in favor of throwing the error such that the user is required to fix it. We want manubot manuscripts to adhere to good standards, especially say for future projects that analyze citation practices.

I'm thinking we need better validation of all identifier types. If we know that a certain ID must adhere to a given format, we should do those checks for the user and provide good error messages, rather than the potentially more confusing error message provided by our retrieval functions.

[agitter]

I'm supportive of throwing an error if the error message is informative. We could potentially make detailed error messages that help catch the most common errors. Something like

PMC6063279 is not a valid PubMed Identifier. You can cite this PubMed Centeral Identifier using pmcid:PMC6063279?