This cleans up the _evaluate method which previously accepted arbitrary args and kwargs, and now requires an expression, a context to which exectute it, and, optionally, the current task being executed if the DMN is being executed as a part of a BusinessRuleTask in a BPMN diagram.
This also cleans up several bits of duplicated code.
There is also a bit of code here to assure that the current user is included when running the master workflow.
1. Avoid ever re-generating the list of scripts that can be used in a script task. Terribly expensive as we call eval constantly, and it never ever changes once the app starts. (see script.py changes, and comments)
2. Cache the DocumentStatus list in the flask session, so we calculate it at most once per API Call. It's at least .25 seconds per call. (see study_sevice)
3. We called UserFileService.get_files_for_study (which runs a db query EVERY time) for every possible document type. Now we run the query once (study service line 321)
4. When returning a workflow, we looped through every single task in that workflow's navigation, and called the expensive spiff_task_to_api_task just to figure out it's proper display name. We run a much faster and more efficient method to calculate the display name naow (see workflow_service on lie 680, and 799)
5. A hellton of @timeit and sincetime() calls, that I want to leave in, to help debug any slowness on production.
it can be kind of irritating for this stuff to be spinning up when you are trying to debug something, so just set
PROCESS_WAITING_TASKS to false in instance/config.py and voila!!
Only load the spec data files if you are creating a new workflow, otherwise just deserialize the json.
Removed the stuff about calculaing the version of the spec, as we don't use it.
Assure we raise more thoughtful error messages when running getting exceptions in engine tasks.
Field Options should always be available now due to a fix in Spiffworkflow.
There was a problem with the python script engine as well that wasn't handling the de-serialize properly and didn't correctly pick back up on the script engine, and the renaming of methods in PythonScriptEngine created some conflicts with the way we override functions.
We were not handling ldap looks up efficiently, and this was also breaking in Study Info.
Finally we had a bug in SpiffWorkflow that did not allow us to reset back to the previous task in some cases where nested call activities happen far later in the process and are currently active when the reset is created.
The problem was that it was returning the first EndEvent it found, not the last one. This caused a problem when we had a CallActivity which has its own EndEvent.
Fixes#399
Most noteable is the addition of the line on which the error occurs for script tasks. It will report the line number and pass back the content of
the line that failed.
The validator only returns the first error it encounters, as it's clear that all we ever get right now is two of the same error.
Did a lot of work between this and spiffworkflow to remove all the places where we obfuscate or drop details as we converted between workflowExceptions and APIExceptions.
Dropped the python levenshtein dependency, in favor of just rolling a simple one ourselves in Spiffworkflow.
We have a test for the function that runs, but an assumption was made that the scheduler module has its own unit tests and we do not need to test that.
fixes#346
