qxlint for VS Code

Deterministic static checks for Qiskit Primitives V2 workflows, in the editor and in notebooks.

qxlint catches the workflow mistakes the Qiskit V1-to-V2 primitives migration introduced: reading counts off the wrong object, using a V1 field on a V2 result, sampling a circuit that has no measurements, and passing a channel value your targeted release has removed.

It never imports or executes your code.

What it catches

Code	Fires when
`QXL101`	`get_counts()` on a `PrimitiveResult`, `PubResult` or `DataBin` instead of the `BitArray`
`QXL102`	`quasi_dists` read from a V2 `PrimitiveResult`
`QXL103`	a provably unmeasured circuit reaches a `SamplerV2`
`QXL104`	a circuit method that returns a new circuit, called as a bare statement
`QXL105`	a measured circuit reaches a `StatevectorEstimator`
`QXL201`	`channel="ibm_quantum"`, removed in `qiskit-ibm-runtime` 0.41
`QXL202`	a Runtime `SamplerV2` or `EstimatorV2` given `backend=` or `session=` instead of `mode=`
`QXL000`	a file or notebook cell that cannot be parsed

Rule codes in the Problems panel link straight to the rule's documentation.

The quiet bug it exists for

An unmeasured circuit sent to a Sampler does not fail. With no classical register Qiskit emits only a UserWarning; with a register but no measure instruction there is no warning at all and every shot reads as zeros, which looks like a physics result rather than a mistake.

Why it does not just match patterns

get_counts() is correct on a BitArray and an AttributeError on a DataBin, and the two look identical in the syntax tree. qxlint resolves what each object actually is, tracks aliases and lists, and stays silent whenever it cannot prove the answer.

Measured on 244 external repositories, 51,715 files: the corpus holds 4,003 .get_counts( calls and 286 quasi_dists occurrences, so a linter matching those textually would have reported 4,289 findings. qxlint reports 18, each on a V2 result object; the rest are correct legacy code.

Requirements

The analyser is the Python package. Install it into the environment you have selected in VS Code:

pip install qxlint

If you have the Python extension installed, qxlint uses the interpreter it has selected and re-runs when you switch environments. It is not required: set qxlint.path to a qxlint executable, or have one on PATH, and the extension works on its own.

Notebooks

.ipynb files are analysed directly, with facts carried across cells in textual order. Magics are classified rather than blanked, because %run can rebind any name: display magics are dropped, %%time bodies are analysed, and namespace-mutating magics act as a barrier so no stale fact survives them.

Settings

Setting	Default	Meaning
`qxlint.enable`	`true`	Run qxlint at all
`qxlint.run`	`onSave`	`onSave` or `onType`
`qxlint.path`	`""`	Explicit executable, otherwise the selected interpreter
`qxlint.select`	`""`	Comma separated codes or prefixes to run
`qxlint.ignore`	`""`	Comma separated codes or prefixes to skip
`qxlint.targetQiskit`	`""`	Target Qiskit version or specifier
`qxlint.targetRuntime`	`""`	Target `qiskit-ibm-runtime` version or specifier
`qxlint.args`	`[]`	Extra CLI arguments

onType re-runs the analyser shortly after you stop typing, but the analyser reads the file from disk, so unsaved edits are not reflected until you save.

Version dependent rules stay silent unless the target version can be established, from these settings or from your project's pyproject.toml. Project level [tool.qxlint] configuration is read by the CLI itself, so the editor and CI agree.

Commands

From the command palette:

Command	What it does
qxlint: Lint workspace	analyse every folder in the workspace, not only open files
qxlint: Re-lint open files	clear the diagnostics and run again, after changing an interpreter or a setting
qxlint: Show output	open the qxlint output channel, which logs the exact command it ran

qxlint

Țugui Dragoș-Constantin