separate exp number and exp name in xml documents

[charlotte]

It will probably make life easier down the line (when connecting questionnaire info with the cmip5 data) if the experiment short name does not also contain the experiment number.

Right now the numbering system ensures that the experiments appear in a logical order in the questionnaire so somehow we need to retain the numbering system for the questionnaire display but not for the questionnaire output.

Can we pull out only the experiment name when we process the questionnaire xml for output?

[spascoe]

Issues to address:

1. How are different CIM records associated with ESG datasets. The THREDDS catalogs of ESG datasets will contain DRS attributes "activity", "institute", "model", "experiment", etc.
2. How to maintain consistent controlled vocabularies between ESG and METAFOR systems. ESG currently uses the CMOR MIP tables as the authoritative vocabulary for experiments. Institutes and models are less clearly defined but exist in Karl's spreadsheets.
3. How will this work both in the ESG Gateway (CIM ingestion problem) and in METAFOR services?

[rupert]

Hi Charlotte,

if we can find a rule (or set of rules) that are guaranteed to distinguish between the experiment number and experiment name then it is trivial for me to strip out the experiment number when we create the associated CIM document

A tentative simple rule is :

The experiment number precedes the experiment name and the first occurrence of a space in the input string acts as the separator between the two.

IMHO the "best" solution would be to input the two bits of information separately into the questionnaire, and for the questionnaire to hold this information separately internally but to concatenate the information in the display. But I understand that this is more work than we need at the moment.

[charlotte]

HI Rupert, yep I can guarantee that all the experiment shortnames follow the same structure so the rule you suggest will work well for the time being.

[rupert]

This is done as of revision 1846. I won't close the ticket yet as

1: rationale and long name also have the id prepended. Do you want these removed too? 2: do we want to keep the id in the CIM output? I'm not sure there is a place for it as things stand. I've just output it as a comment for the moment. 3: Stephens questions have not been answered although I think they are much more broad than the original intent of this ticket.

[charlotte]

Hi Rupert, pedantic point but the number associated with each experiment is not strictly speaking an "id" because it is not unique. Would be better to call it an experiment "number". As for whether we want to keep the number in the CIM output... I think it needs to be in the output somewhere for although the experiment numbers are not used by the DRS they are an integral part of the experiment descriptions in the CMIP5 experiment design document. I suggest we prepend the description with the experiment number in the CIM output.

[rupert]

Hi Charlotte,

I am happy to add it to the description, however that will effectively make it impossible to be part of a targeted search. I have an alternative suggestion which might conflict with your pedantic point :-). The CIM has particular properties associated with being an experiment. One of these is an experimentID. This is unused at the moment in the q2cim translation. It seems to me that this would be a natural place to put this. I will do this unless anyone shouts loudly. I Cc'ed Allyn in case he has a view. Similarly, I will remove the experiment number from the long name and the rationale unless you say otherwise.

[allyn]

This is overloading experimentID; it was meant to be a unique identifier used by the community. It sounds like the information you want to record is specific to the CMIP5 Questionnaire. If so, then I can add an optional experimentNumber attribute for this purpose.

[charlotte]

I'm pro adding an optional experimentNumber, this would make sense to CMIP5 users as a search term too.
As for the unused experimentID... the experiment shortname is very much like an id for CMIP5, each is unique so that the experiments can be identified by software that knows about DRS syntax. It would make more sense to use the short name as the id for the CMIP5 experiments.
I don't think it is necessary remove the experiment number from the long name and rationale because the number gives a context for each experiment.

[allyn]

As of r1852, I have removed experimentID and added experimentNumber.

[rupert]

As of revision 1853 the experiment number is now output in the new experimentNumber element. In addition the experiment number is also stripped from the shortName, longName and rationale. I am therefore closing this ticket.