Support (de)serializing yaml-format recipes

[rjgildea]

This works by parsing serialized recipes with a YAML 1.2-compatible parser, of which JSON is a strict superset.

The following JSON-format recipe:

{
  "1": [
    "Example recipe for per-image-analysis storing results in ISPyB directly"
  ],
  "1": { "service": "DLS file watcher",
         "queue": "filewatcher",
         "parameters": { "pattern": "/path/to/images/2A4_1_%04d.cbf",
                         "pattern-start": "100",
                         "pattern-end": "100"
                       },
         "output": { "every": 2 }
       },
  "2": { "service": "DLS Per-Image-Analysis",
         "queue": "per_image_analysis",
         "output": 6
       },
  "6": { "service": "DLS ISPyB connector",
         "queue": "ispyb_pia",
         "parameters": { "ispyb_command": "store_per_image_analysis_results",
                         "program_id": "$ispyb_autoprocprogram_id",
                         "dcid": "{ispyb_dcid}"
                       }
       },
  "start": [
      [1, []]
  ]
}

could equivalently be represented in YAML format as:

1:
  output:
    every: 2
  parameters:
    pattern: /path/to/images/2A4_1_%04d.cbf
    pattern-end: 100
    pattern-start: 100
  queue: filewatcher
  service: DLS file watcher
2:
  output:
    - 6
  queue: per_image_analysis
  service: DLS Per-Image-Analysis
6:
  parameters:
    dcid: {ispyb_dcid}
    ispyb_command: store_per_image_analysis_results
    program_id: $ispyb_autoprocprogram_id
  queue: ispyb_pia
  service: DLS ISPyB connector
start:
- [1, []]

workflows.recipe.Recipe.serialize() now supports the keyword argument format="json". Acceptable values are "json" or "yaml".

No performance comparisons between the two parsers have been made yet. It could be possible to wrap the parsing in a try/except whereby we first try parsing with the native JSON parser, and failover to YAML in the event of failure, although this does complicate error handling.