WIP: MDS Spec versus MDS Implementation

NSLS-II / docs

Standards Documentation for NSLS-II DAQ and Analysis

https://nsls-ii.github.io

BSD 2-Clause "Simplified" License

2 stars 13 forks source link

WIP: MDS Spec versus MDS Implementation #2

Closed ericdill closed 9 years ago

ericdill commented 9 years ago

based on conversations with @danielballan, @tacaswell, @arkilic What other changes do we need guys?

ericdill commented 9 years ago

iirc the event_descriptor does not have a type key anymore, right?

ericdill commented 9 years ago

oh, meant to loop @dchabot in on this too, sorry!

stuwilkins commented 9 years ago

Hi Eric,

If you look at the top of the document it explains why types are required. While the trigger has gone (I am still a little uncomfortable about that, but @dchabot convinced me) One could think about putting this in the top run header, but I believe this reduces flexibility.

I would suggest not making changes without the small group getting together. @tacaswell made some really good points about how the current metadatastore can be used for different “stages” of data and building in that flexibility at this stage is really important.

Please call a meeting of the few if you would like to change things.

Thanks, Stuart

On Jan 28, 2015, at 5:38 PM, Eric Dill notifications@github.com<mailto:notifications@github.com> wrote:

iirc the event_descriptor does not have a type key anymore, right?

— Reply to this email directly or view it on GitHubhttps://github.com/NSLS-II/docs/pull/2#issuecomment-71931893.

dchabot commented 9 years ago

On Wed, Jan 28, 2015 at 5:38 PM, Eric Dill notifications@github.com wrote:

oh, meant to loop @dchabot https://github.com/dchabot in on this too, sorry!

I've been following. But, I do not recall when the 'type' key was scheduled for execution. Was it ever included to begin with? Is it of any use 'downstream' (i.e. clients)?

ghost commented 9 years ago

It was never part of the descriptor.

On Jan 28, 2015, at 6:03 PM, dchabot notifications@github.com<mailto:notifications@github.com> wrote:

On Wed, Jan 28, 2015 at 5:38 PM, Eric Dill notifications@github.com<mailto:notifications@github.com> wrote:

oh, meant to loop @dchabot https://github.com/dchabot in on this too, sorry!

I've been following. But, I do not recall when the 'type' key was scheduled for execution. Was it ever included to begin with? Is it of any use 'downstream' (i.e. clients)?

— Reply to this email directly or view it on GitHubhttps://github.com/NSLS-II/docs/pull/2#issuecomment-71935560.

danielballan commented 9 years ago

I am sitting beside Arman as he finishes implementing this. Here are some decisions we made that deviate from the documented spec:

Event descriptors are documented to have keys, but this clobbers the Python method, so we are reverting to data_keys. We don't remember any discussion on this point anyway.
Sample is currently implemented as a simple dict, not a full-blown collection like Beamline Config. Perhaps in the medium-term it should be an DynamicEmbeddedDocument.
Begin Run Events are indexed on _id, uid, time, owner, and scan_id. End Run Events are indexed on _id, time, exit_status, and begin_run_event.
Arman prefers the name exit_status to completion_state and I agree. Can revert later if others dissent.

tacaswell commented 9 years ago

These specs shouldn't say anything about indexing

tacaswell commented 9 years ago

also, work against #3 not this one