The `mesh_report_parsers.py` module#

Summary#

Functions

`parse_mesh_quality`	Extract the three headline quality numbers from a `mesh.quality` report.
`parse_mesh_check`	Extract structured fields from a `mesh.check` report.

Description#

Pure-functional parsers for Fluent’s mesh.check and mesh.quality reports.

These commands print a human-readable block to the Fluent transcript/console. PyFluent forwards that print stream to Python’s stdout, which ansys.fluent.mcp.solve.backends.pyfluent.PyFluentBackend.run_code() captures into RunCodeResult.stdout via contextlib.redirect_stdout. The following parsers take that captured text and extract structured numbers/lists.

Why parse instead of using a programmatic API:

mesh.check and mesh.quality are settings-API commands. They return None and emit their result to the console. There is no structured return value across the supported Fluent versions. (Newer PyFluent versions have experimental structured reports under solver.report.*, but the names have churned across the 24.x / 25.x) releases.
The transcript wording for the headline lines cared about (Minimum Orthogonal Quality/Maximum Ortho Skew/Maximum Aspect Ratio/Volume statistics/Face area statistics/Domain Extents) has been stable since Fluent 19.0. Extra histograms/hints below the headline lines are version-dependent and ignored.

Fluent 2024 and later formats notes (verified against a real session):

mesh.quality typically prints only Minimum Orthogonal Quality and Maximum Aspect Ratio. Maximum Ortho Skew is often absent because Fluent treats orthogonal quality as the canonical cell-quality metric (orthogonal quality ≈ 1 − ortho-skew, so the two are redundant). The parser returns None for the missing field. Callers MUST treat None as unknown and rely on min_orthogonal_quality as the primary signal.
The headline lines now include trailing annotations like cell on zone (ID: on partition:
) at location (x, y, z). These are ignored by design. Only the number to the right of the = is captured.
mesh.check may indent the Volume statistics: and Face area statistics: headers with one leading space and use single-space separators between total volume and (m3). The regex whitespace classes (\\s+, \\s*) tolerate both styles.

Both parsers are intentionally lenient. Anything that cannot be matchd collapses to None rather than raising. Callers are expected to treat None as unknown, never as a passing score.

Tested against transcripts from both pressure-based and density-based solvers in tests.test_mesh_report_parsers.

Module detail#

mesh_report_parsers.parse_mesh_quality(stdout: str | None) → dict[str, float | None]#

Extract the three headline quality numbers from a mesh.quality report.

The returned mapping contains "min_orthogonal_quality", "max_ortho_skew", and "max_aspect_ratio". A value is None when the line is not found in stdout (no mesh loaded, command failed, transcript capture truncated). Never raises.

Parameters:

stdoutstr | None: Stdout to supply to the function.

Returns:

dict[str, float | None]: Parsed quality metrics keyed by Fluent’s headline mesh-quality fields.

mesh_report_parsers.parse_mesh_check(stdout: str | None) → dict[str, Any]#

Extract structured fields from a mesh.check report.

Output shape is documented on ansys.fluent.mcp.common.backend.Backend.mesh_check(). Missing fields are emitted as None (or empty list for warnings/errors). Never raises.

The raw field always carries the verbatim stdout chunk so callers that need a feature this parser doesn’t extract yet can still get at it without re-running the command.