Add architecture design document #41
Conversation
| @@ -0,0 +1,9 @@ | |||
| # Architecture and Design | |||
There was a problem hiding this comment.
Why not "Architecture decision records"? This would make it easier to find given the Scipp docs and it allows us to add more if need be.
There was a problem hiding this comment.
I had considered adding this as a subsection. I thought we might have other documentation here as well, not just decision records?
| transmission_monitor: sc.DataArray, | ||
| """ |
There was a problem hiding this comment.
The docstring has slipped in between the arguments and return annotation.
| However, instead of the user having to pass these to functions, we use dependency injection to provide them to the functions. | ||
| In essence this will build a workflow's task graph. | ||
|
|
||
| From the [Guice documentation](https://github.com/google/guice/wiki/MentalModel#injection) (Guice is a dependency injection framework for Java): |
There was a problem hiding this comment.
Move to the Dependency injection section.
| incident_monitor: IncidentMonitor, | ||
| transmission_monitor: TransmissionMonitor, | ||
| ) -> TransmissionFraction: | ||
| return transmission_monitor / incident_monitor |
There was a problem hiding this comment.
| return transmission_monitor / incident_monitor | |
| return TransmissionFraction(transmission_monitor / incident_monitor) |
It's unfortunately verbose but should be part of the design discussion.
| Generally, the user must provide configuration parameters to a workflow. | ||
| In many cases there are defaults that can be used. | ||
| In either case, these parameters must be associated with the correct step in the workflow. | ||
| This is complicated by the non-linear nature of the workflow. |
There was a problem hiding this comment.
| This is complicated by the non-linear nature of the workflow. | |
| This gets complicated by the non-linear nature of the workflow. |
?
| ## Notes | ||
|
|
||
| - Earlier versions of this design document included a detailed discussion on how to handle nested workflows using child-injectors. | ||
| During initial implementation efforts this idea was discarded, as it was found to be too complicated and not necessary. |
There was a problem hiding this comment.
Um, what is duplicate?
There was a problem hiding this comment.
Maybe not a full duplicate. But the other section also mentions that nested workflows were considered but not implemented.
There was a problem hiding this comment.
Isn't the section talking about nested parameters, as opposed to nested workflows?
Fixes scipp/ess#157.