Refactor to align with changes in databroker about csxtools

[ambarb]

unsure if this particular instance was ever fixed / addressed. @stuwilkins let me know if the changes below do not address the issue that started this thread

Taking over the issue for the current problems. Using get_table() because .data() does not work with current version of databroker at CSX.

Noting that we used databroker 0.13.0 doesn't play nice with the default HDF5 handler for area-detector. Only get_table() seems to work (which is why CSXtools still uses it. Testing with @danielballan confirmed handler issue. However, the issue is not present in databroker 1.0.2

@leofang FYI, but I don't think we are ready to make the switch yet, just thought you would be interested.

[danielballan]

As far as you know, is csxtools broken by our API changes, invoking our deprecations, or none of the above?

[stuwilkins]

All of the above

[danielballan]

OK. If this is disruptive to user operations and you lack the bandwidth to address it in time, I'm willing to help.

[stuwilkins]

@danielballan don't worry, I think it is good that I (we) have to change this. I will work on this from home wed/thurs to get all inline. I may need help with handlers etc? Where is the idea of adding a config file for defining how the handlers get assigned to specs? That is by biggest need at the moment in this respect (CSXTOOLS)

[stuwilkins]

Oh and we do need to get FastCCD data into databroker correctly....

[danielballan]

Where is the idea of adding a config file for defining how the handlers get assigned to specs?

Nominally implemented and deployed, but when I looked up the relevant code to back up that claim, I found that it's undocumented, untested, and broken. All those problems are addressed in https://github.com/NSLS-II/databroker/pull/305, which will be merged and deployed at CSX today.

[danielballan]

Oh and we do need to get FastCCD data into databroker correctly....

With regard to the timing of when documents are created?

[danielballan]

@stuwilkins: @danielballan don't worry, I think it is good that I (we) have to change this. I will work on this from home wed/thurs to get all inline.

csxtools still has a filestore import, which made it un-importable in the 2017-3 conda environment. It would be nice to fix it in time for 2018-1. Do you need help with that?

[danielballan]

attn @wen-hu

[danielballan]

@stuwilkins I looked through all the branches on your fork, and I don't see one that addresses the subject of this issue. @ambarb mentioned there might be one on srv3. Any guidance?

[ambarb]

stale

addressed also with #86