ServiceX Helper
A VS Code extension for managing ServiceX
transform caches and jobs.
It adds a "ServiceX Helper" icon to the Activity Bar with a "Cached Transforms" view,
which lists every transform request in your local ServiceX cache, grouped by
title and refreshed from the backend.
Hovering over a row gives you:
- On a single request: Delete from Cache - removes its downloaded files
and cache record.
- On a title group: Clean Old Requests - keeps only the most recently
submitted completed request under that title, deleting older completed
runs plus any cancelled or failed ones, or Delete Entire Group -
removes every locally cached request under that title.
Right-clicking a single request also gives you:
- Copy Request ID - copies the request ID to the clipboard.
- Copy File List - copies the paths of that request's downloaded files
(one per line) to the clipboard.
- Show Query - opens the query that produced the request in a scratch
editor. See below.
Show Query
The local cache doesn't keep your query - only the hash it was reduced to -
so Show Query asks the backend for the transform's selection string and
decodes it. What comes back depends on how you submitted the request:
| Query type |
Stored as |
You get back |
FuncADLQuery |
qastle (an s-expression form of the Python AST) |
Equivalent func_adl source |
PythonFunction |
base64 of the function's source |
That source, exactly |
UprootRawQuery |
JSON |
That JSON, exactly |
TopCPQuery |
JSON with the YAML inlined |
That JSON, exactly |
Only func_adl queries are approximate: qastle preserves the query's
structure, its lambda parameter names, and every literal, so the recovered
source runs the same way - but comments, formatting, and the names of any
intermediate variables you built the query out of aren't in there to recover.
Two things happen to a func_adl query on the way out. The generated
MetaData(...) wrappers are removed: the backend packages
(func_adl_servicex_xaodr25 and friends) add one for every C++ type and
method the query touches - include files, link libraries, injected code - and
a query with a couple of dozen .pt() calls can end up under seventy of
them, which buries the Selects you actually wrote. Set
servicex.showQueryMetadata to keep them. The result is then laid out over
several lines, using black if the interpreter
has it and a built-in fallback formatter otherwise.
Decoding qastle needs a Python interpreter with the qastle package, since
there's no JavaScript implementation of it. Anyone who has submitted a
func_adl query already has one (servicex depends on func_adl, which
depends on qastle), and the extension looks for it in this order:
servicex.pythonPath, if you set it.- The interpreter the Python extension has selected for the workspace.
python3, then python, from PATH.
If none of those work out, the raw qastle is shown instead, indented for
readability, with a note explaining what to fix. Over Remote-SSH all of this
runs on the remote, alongside the environment you submit queries from.
Development
npm install
Then open this folder in VS Code and press F5 to launch an Extension
Development Host with the extension loaded.
Testing
npm test
Runs the full test suite inside a real (headless) VS Code instance via
@vscode/test-cli. Also
runs automatically on every push and pull request via
.github/workflows/test.yml.
License
BSD 3-Clause
