Output settings and publication
Root outputs selects result table formats, per-message contributions, and additional data exports. AnalysisRunner.run() performs the analysis and prepares staged output; result.write(output_directory) publishes it. CLI run performs both steps.
Analysis tables describe what happened and the calculated values. Signal CSV files contain timestamps and values. ASC/BLF files contain CAN message bytes. These outputs serve different purposes and can be requested together. Input and output formats need not match: ASC input can be exported as BLF.
Resampling produces values at regularly spaced times you specify. For example, if voltage is 10 V at 10 ms and 120 V at 120 ms, an output every 100 ms can estimate 100 V at 100 ms along the straight line between those observations. The output marks that value estimated; it is not an actual sample at 100 ms. Because it uses the later 120 ms observation, this is an offline operation. Rules and integration still use the original data, not these estimates.
Device labels and output locations
metadata.device_label records the device associated with this run. It is optional and defaults to pending. To record a device, supply any nonblank string, such as lab-pc or 试验台一号. The same rule applies to ordinary analysis, signal CSV exports, and CAN frame exports. Labels do not affect calculations. Empty, whitespace-only, or non-string values fail configuration validation.
Choose your output directory freely. Existing result directories can be reused: each run receives new filenames and preserves existing files. An explicit working directory must still be new. Ordinary CLI analysis does not override this label from the environment; see Batch tasks for the batch report label.
All outputs fields
outputs is an optional root object, for example outputs: {tables: jsonl}. Paths in this table are relative to it: tables means outputs.tables. Table format does not change the calculation, and the two additional export types can be enabled independently.
| Field | Type | Required | Default | Choices and effect |
|---|---|---|---|---|
tables |
String | No | "readable_csv" |
"readable_csv" creates readable Chinese tables joining related intervals and metrics; "csv" preserves the flat exchange fields; "jsonl" writes corresponding summary rows as line-delimited JSON. Select the format appropriate to the receiving tool. |
manifest |
String | No | "json" |
Run manifest format; use json or omit it. The manifest describes this run's inputs, settings, and outputs. |
contributions |
String | No | "none" |
"full" retains each target message's evaluation and skip details. "none" accumulates summaries without retaining per-evaluation detail counts. This switch applies to per-message calculations, not ordinary calculations. |
signal_csv |
Object or null |
No | null, disabled |
Enable selected original-signal export, optionally with resampling. See Signal exports. |
can_frames |
Object or null |
No | null, disabled |
Enable frame export from one CAN source, choosing format and optional reconstruction. See CAN exports. |
BLF, ASC, and raw-message CSV frame exports all use outputs.can_frames.
Export scope
scope value |
Effect | Supported exports |
|---|---|---|
Omitted or session |
Export the session range. | Signal CSV and CAN frames |
{window: context} |
Export segments from the window named context. |
Signal CSV and CAN frames |
{windows: all} |
Export segments from every window. | CAN frames only |
Window names come from windows[].name. Set the choice in outputs.signal_csv.scope or outputs.can_frames.scope; do not add a separate window field. null is not a valid scope. Signal export may omit scope for the session range. Specify one named window only when selecting window-based export; signal export does not accept windows: all.
Each export chooses its own scope independently. Ranges include the start and exclude the end: a 1–2 second segment includes 1 second but not 2 seconds. Interpolation does not borrow samples from outside the requested range to fill endpoints.
Choose an export type
- Signal CSV and resampling: export physical-value updates or regularly spaced values.
- CAN frames and resampling: export original ASC/BLF frames or reconstruct selected fields.
Find the output files
run first prepares exports in the work directory. After result.write(), they appear beside the analysis tables in the final output directory.
Read the timestamped index named by the run manifest's exports.index_file there to find each export's time range, source, and file paths. Data files are under exports_<timestamp>/export-XXXXXXXX/. Content digests are retained only with diagnostics.integrity: true; default checks cover necessary file identity and size.
Writing the same open result to the same or another output directory checks and copies already generated files; it does not reread or decode the original logs.
The accompanying example holds 3 V, 7 V, and 3 V for one second each, giving 13 V·s, with voltage above 5 V over [1,2). Requesting a 250 ms signal grid and a 500 ms CAN grid together leaves those analysis results unchanged. See Examples, especially the dual-grid example.
Full classes, functions, and arguments are available in signal_exports, signal_grid, frame_exports, and frame_rebuild (Chinese implementation references).
Direct Python calls to export_signal_csv or export_can_frames return ExportManifest with output_dir, manifest_path and output_files. The directory is reusable and files are timestamped. Read these properties instead of treating the return value as a Path. Exports requested through a plan are located through the run manifest exports entry.