Possibly have to use workspace/didChangeWatchedFiles and allied mechanisms.

Allows to quickly navigate to important sections of the document. Proposed sections are

1. inputs
2. inputs-end
3. outputs
4. outputs-end
5. steps
6. steps-end

After the initial parse an edits we do can possibly be incrementally added on to the original parse faster than re-parsing, but this will be more complex to do.

Probably relates to #17

Parse the $namespaces field and use that to parse namespaced values.

We should store $namespaces in the Code Intelligence Object and use this just as we use user types to validate class names. For the first version we won't validate the actual class definition

This is a better fit (rather than have it share space with the connection table).

1. create wf
2. new step1
3. create wf
4. new step1
5. create clt
6. Remove trailing endline. The endline is programmatically added back no particular op is needed. Bug exists regardless.
7. Switch to middle tab. Workflow is not updated

dukpy sounds promising

While technically correct, it is irritating to clear out a run field in a nested editor and have it be re-cast as a text block in the parent. We should alter the logic in YamlDoc to make exceptions for specific fields.

Instead of just a test, add a chip like schematic showing all inputs and outputs of the CLT or ET.

Otherwise it is hard to find/navigate to issues in a workflow with many lines

cwltool does it this way https://github.com/common-workflow-language/cwltool/blob/03af16c9df2ee77485d4ab092cd64ae096d2e71c/cwltool/main.py#L188

Select a set of step ids and group them into a new sub-workflow. Recompute all connections as needed

These can be written better now that we understand the problem domain more fully.

e.g.

requirements:
  - class: ScatterFeatureRequirement
  - class: InlineJavascriptRequirement
  - pass: MultipleInputFeatureRequirement

will flag the error on the first item of the list, rather than the last

Need to protect if p[-1] == "run" and p[-3] == "steps": from
IndexError: tuple index out of range errors

https://doc.qt.io/qtforpython/overviews/qtest-overview.html#qt-test-overview

https://pytest-qt.readthedocs.io/en/latest/

when trying to edit https://github.com/EBI-Metagenomics/workflow-is-cwl/blob/master/workflows/TranscriptomeAssembly-wf.paired-end.cwl the editor will crash with the following tb:

INFO:benten.langserver.server:Client query: textDocument/didOpen
ERROR:benten.langserver.server:Unexpected error: unhashable type: 'CommentedSeq'
Traceback (most recent call last):
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/langserver/server.py", line 63, in run
    self.handle(request)
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/langserver/server.py", line 93, in handle
    response = to_dict(self._dispatch(client_query))
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/langserver/server.py", line 142, in _dispatch
    return f(client_query)
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/langserver/fileoperation.py", line 60, in serve_textDocument_didOpen
    language_models=self.config.lang_models)
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/models/document.py", line 87, in __init__
    self.update(text)
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/models/document.py", line 96, in update
    parse_document(cwl, self.doc_uri, self.text, self.language_models, problems)
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/models/languagemodel.py", line 221, in parse_document
    requirements=None)
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/models/languagemodel.py", line 824, in parse
    completer.wf_completer.analyze_workflow(node, doc_uri, problems)
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/models/languagemodel.py", line 927, in analyze_workflow
    self.validate_step_connections(step, step_id, problems)
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/models/languagemodel.py", line 976, in validate_step_connections
    if conn_id not in self.all_connections:
TypeError: unhashable type: 'CommentedSeq'
INFO:benten.langserver.server:Client query: textDocument/documentSymbol
ERROR:benten.langserver.server:Unexpected error: 'file:///home/hmenager/galaxycwl/workflow-is-cwl/workflows/TranscriptomeAssembly-wf.paired-end.cwl'
Traceback (most recent call last):
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/langserver/server.py", line 63, in run
    self.handle(request)
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/langserver/server.py", line 93, in handle
    response = to_dict(self._dispatch(client_query))
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/langserver/server.py", line 142, in _dispatch
    return f(client_query)
  File "/home/hmenager/.local/pipx/venvs/benten/lib/python3.7/site-packages/benten/langserver/documentsymbol.py", line 22, in serve_textDocument_documentSymbol
    doc = self.open_documents[doc_uri]
KeyError: 'file:///home/hmenager/galaxycwl/workflow-is-cwl/workflows/TranscriptomeAssembly-wf.paired-end.cwl'

When we have an issue with a parse, we use the last known good model, which is fine. But we should also pass on the issues with the current parse so the user can act on them.

• Feature to create an input JSON from the CWL #49
• Mechanism to evaluate JS #50

Somewhere in our protocol we are probably missing edits that follow rapidly

Activated on CTRL+F or other shortcut, unobtrusive, simple, no reg-exp

Server should offer option to rename ports and steps which will propagate within the given file.

Out of scope: renaming port in a workspace.

https://microsoft.github.io/language-server-protocol/specification#textDocument_rename

[ ] Think about log/error handling

Tab out whole left pane, tab out just the connection table?

Examples should include those found in #39

This is a little dialog that allows you to select which labels will always be shown next to a step on a workflow (like annotation layers on a map).

This class is getting big. It should be moved to it's own directory and divided out between a base class and a series of mixins designed to make adding new functionality more principled and modular.

When the plugin starts up we should check if benten-ls exists, if not, it should try and install it.

This is the reverse of clicking on a map node and scrolling the code editor to that line.

The documentation is fine but could be improved:

• add a "quick setup" for various setups, like ubuntu+vscode (stub here: https://gist.github.com/hmenager/2c56c3fc6ab25c32ad03d4becdabd4bb)
• the "VI setup" page should be restructured in a more chronological "step-by-step" order.

Let's see if this is needed, or it's easy enough to scroll and find the end.

Can occasionally be spit out when closing a session with multiple tabs. Should track this down eventually.

The new token based parser will handle this as it will add marks to every thing including keys.

Took it out during the refactor, need to put it back in again. ~30 min worth of work.

1. Reduce the number of YAML parses
2. Restrict YAML parses to on demand

This needs addition on the server side and the VS code client and new code for the grammar

Hi, just started using Benten and it's awesome! Just noticed a few small things I'm no sure if my CWL is actually valid or Benten isn't recognising something properly.

I have a Workflow Input of type, but I've got a yellow squiggle under items (Unknown field: items for type InputRecordSchema) and red under array (Expecting one of: {'record'}):

  fastq:
    id: fastq
    type:
      items: File
      type: array

Using that latest development per: pipx install --spec git+https://github.com/rabix/benten.git@develop benten

Also, this line on the README is missing the second hyphen before spec (--)

The Pyzo editor looks pretty modular and could be adapted for our needs. This would provide a nicer code editor pane.

This was only really needed for when we had our own editor. As a lang server this is not needed any more - there are no lang server options there, except for autocomplete templates and that we can probably fix by having the template located in a fixed place with a fixed name. This auto complete templates file can probably also carry hints autocompletions. Only these items are user dependent and should go in a config file.

Currently strings in the CWL that represent paths are treated as system dependent paths. This has to be fixed to treat these paths as POSIX paths and then translated in a system dependent manner. Windows also requires proper resolution of paths like

C:\\My\Path\..\path2

which seems to happen fine on macOS