Add nextstrain run support

[joverlee521]

Description of proposed changes

The goal of the repo is to present the simplest version of a phylogenetic workflow so I have kept the top level Snakefile and kept the original behavior where the input files are hard-coded in the workflow and cannot be modified via config parameters.

Requires Nextstrain CLI 10.4.0 with support for the workflow name to path mapping to be able to run with

nextstrain setup zika-tutorial@nextstrain-run-support
nextstrain run zika-tutorial@nextstrain-run-support phylogenetic /tmp/zika-tutorial

Related issue(s)

Resolves #18

[tsibley]

Since the Snakefile is in the top level directory, use . as the workflow name for nextstrain run:

That is a very weird name for a workflow, esp. a workflow we admit is "the simplest version of a phylogenetic workflow" (emphasis mine).

How about we use this opportunity to implement workflow name → directory mapping in nextstrain-pathogen.yaml? e.g. something like:

workflows:
  phylogenetic:
    path: .

which would allow:

nextstrain run zika-tutorial@nextstrain-run-support phylogenetic /tmp/zika-tutorial

Or, alternatively, we could move everything into phylogenetic/ and call it a day for now.

[tsibley]

How about we use this opportunity to implement workflow name → directory mapping in nextstrain-pathogen.yaml?

If we want this, it's pretty easy to support:

diff --git a/nextstrain/cli/pathogens.py b/nextstrain/cli/pathogens.py
index 5dd0172..411c574 100644
--- a/nextstrain/cli/pathogens.py
+++ b/nextstrain/cli/pathogens.py
@@ -342,8 +342,19 @@ def compatible_workflows(self, feature: str) -> Dict[str, Dict]:
         }
 
 
-    def workflow_path(self, workflow: str) -> Path:
-        return self.path / workflow
+    def workflow_registration(self, workflow_name: str) -> Optional[dict]:
+        if (info := self.registered_workflows().get(workflow_name)) and not isinstance(info, dict):
+            debug(f"pathogen registration.workflows[{workflow_name!r}] is not a dict (got a {type(info).__name__})")
+            return None
+        return info
+
+
+    def workflow_path(self, workflow_name: str) -> Path:
+        if (info := self.workflow_registration(workflow_name)) and (workflow_path := info.get("path")):
+            return self.path / workflow_path
+
+        debug(f"pathogen registration does not specify path for workflow {workflow_name!r}; using name as path")
+        return self.path / workflow_name
 
 
     def setup(self, dry_run: bool = False, force: bool = False) -> SetupStatus:
@@ -747,6 +758,7 @@ def __init__(self, path: str):
 
     registered_workflows = PathogenVersion.registered_workflows
     compatible_workflows = PathogenVersion.compatible_workflows
+    workflow_registration = PathogenVersion.workflow_registration
     workflow_path = PathogenVersion.workflow_path
 
     def __str__(self) -> str:

[tsibley]

(There's some safeguards we'd want to add, like ensuring the registration-provided path is relative and does not specify a location outside the containing dir.)

[joverlee521]

That is a very weird name for a workflow, esp. a workflow we admit is "the simplest version of a phylogenetic workflow" (emphasis mine).

Hah, definitely agree! I wasn't happy with the . workflow name hence the draft PR.

How about we use this opportunity to implement workflow name → directory mapping in nextstrain-pathogen.yaml

+1 for workflow name to directory mapping to solve this weirdness.

[tsibley]

The release of Nextstrain CLI 10.4.0 yesterday adds support for the workflow name → path mapping.