ChronaxisUser manual中文
On this pageRun the basic parallel configurationCommandsExit codes and failure stagesSee several configuration errors at onceAI guide

Command-line reference

Fast single-session commands fast validate, fast explain and fast run accept --process-count to override analysis execution.process_count. fast batch --process-count overrides its batch-root count. One fast session can contain multiple ASC/BLF volumes and publishes one summary. Workers read, decode and evaluate current-message contributions; the main process accumulates in order. See fast configuration.

Use terminal commands to load and check a configuration, run an analysis, and publish results. Put the analysis definitions in the configuration file; put options such as an output directory or backend override after the appropriate command.

Run the basic parallel configuration

Download the complete examples directory and open a terminal there. These steps inherit the continuous-volume example's explicit execution.process_count: 4:

text
python -m chronaxis validate continuous-parallel.yaml
python -m chronaxis explain continuous-parallel.yaml --format json
python -m chronaxis run continuous-parallel.yaml --output result-continuous

validate checks configuration, explain describes execution, and run reads and analyzes the logs. With two files, source_execution[].file_count and read_process_limit are both 2; four is only the limit. Use manifest in the completion summary to locate this run's actual output files. The integral is 38 J; see the worked result.

Ordinary explain reports read_process_limit as a bound derived from configuration and source type, not the number of workers already started. For one ASC, configured limit 4 and default concatenate, reading takes place in the current process and read_process_limit is 0; two files with that limit yield 2. Ordinary time_order uses one file child even at a configured limit of 1. Fast worker_process_limit follows its own scheduling rules and is 0 for inline execution. These are plan facts; actual PIDs can only be established during execution.

Append --process-count 2 for a temporary override without editing the configuration. A single CSV, memory or multi-source plan uses process_count: 1. Independent tasks use the batch-root count; see execution settings.

Commands

Command Arguments and options Purpose
python -m chronaxis config init/convert/resolve/origins See Configuration tools and Shared definitions. Generate templates, convert formats, expand configurations, or inspect declared origins.
python -m chronaxis validate PLAN Optional --backend auto/reference/numpy, --process-count INTEGER, --format text/json. Check the plan; readable text is the default, JSON is available for tools.
python -m chronaxis explain PLAN Same options as validate. List required signals, resource limits, input-check scope, and calculation implementation.
python -m chronaxis run PLAN Required --output; optional --work-dir, --backend, --process-count. Analyze, publish, close, then print a completion summary.
python -m chronaxis batch run BATCH Required --output; optional --process-count. Run independent plans and write a per-job report; see Batch execution.
python -m chronaxis fast validate PLAN Plan path Check the supported same-frame calculation and DBC bindings; prints JSON.
python -m chronaxis fast explain PLAN Plan path Show the input file, calculation definitions and failure policy as JSON.
python -m chronaxis fast run PLAN Required --output; optional --process-count Compute one session from its ASC/BLF volumes and publish a summary.
python -m chronaxis fast batch BATCH Required --output; optional --total and --process-count Compute independent files. --total requests totals for comparable template calculations; see fast batch.
python -m chronaxis guide Optional --language zh/en, --output new.zip Print the existing README path; save a bilingual ZIP only if output is requested.

Replace PLAN and BATCH with paths. An installed package also provides the chronaxis command. Success returns 0 and cancellation returns 130. Failure codes 1 and 2 depend on the command and stage, as explained in Exit codes and failure stages; “file error” alone does not determine the code. Use python -m chronaxis --help and each subcommand's --help. The full implementation is in the CLI reference (Chinese implementation reference).

Ordinary run, validate and explain accept --process-count to override file-reading/decoding workers. The same option for batch run limits independent analysis workers. See continuous multiprocessing.

For fast run, execution.process_count controls session workers; fast batch uses the batch-root count. Both accept --process-count overrides, as do fast validate/explain. --total is a flag, omitted by default; the report explains any unavailable total. For guide, --language defaults to zh; en selects the English README, and an optional ZIP contains both languages.

Argument or option Type Required for Default when omitted Accepted values and effect
PLAN File path validate, explain, run None Analysis YAML/JSON path; quote paths containing spaces.
BATCH File path batch run None Batch YAML/JSON, not one child analysis.json.
--format String Optional for validate/explain text text prints readable diagnostics; json returns a report object. It does not change result table formats.
--format String Optional for config resolve Terminal: yaml; file: suffix Expanded configuration as yaml/json; an explicit value must match the output suffix.
--format String Optional for config origins text text lists origins; json gives a field-origin report, not analysis results.
--backend String Optional Plan's backend auto selects from the whole plan; reference and numpy fix the numerical implementation. This option overrides the plan. Applies to validate, explain, and run.
--output Directory path run, batch run, fast run/batch Required Reusable result directory; each run creates distinct timestamped files.
--work-dir Directory path Optional; ordinary run only Create a dedicated work directory Holds intermediate data and staged results. An explicit path must not already exist.
--process-count Integer Optional; run/validate/explain/batch run and fast equivalents Ordinary analysis: execution.process_count; batch: root process_count; both default 1 Positive integer. Ordinary analysis limits read/decode workers; fast session workers also evaluate contributions. Both exclude the main process. Batch limits independent analysis workers, excluding their existing CAN subprocesses.
-h / --help Flag without a value Optional Execute the command Display help for that level and exit without analysis, e.g. python -m chronaxis run --help.

For config init, convert, and resolve options and file-protection rules, see Configuration tools. For direct reading and optional guide export, see AI guide.

Exit codes and failure stages

The exit code tells you whether a command succeeded; diagnostics and batch reports explain why it failed. The same file problem may return different codes through different commands. Account for the command when writing a shell script.

Command Common failures returning 2 Common failures returning 1
Ordinary validate, explain, run Invalid arguments or configuration, an empty source directory, or a glob with no matches Unreadable configuration or DBC; for run, log reading, analysis, publication or close failures
Ordinary batch run Invalid batch configuration or arguments; configuration errors while expanding template/inputs or constructing child plans during loading Unreadable batch or template configuration files; a jobs child configuration rejected after its task starts, or another failure in a started task; output, batch-report or cleanup failures
fast validate, fast explain, fast run Configuration, DBC or input-list preparation failures, including file errors at those stages For fast run only: log reading, calculation or result publication failures after preparation
fast batch Batch preflight or output-directory setup failures; a final batch-report write failure without a recoverable report attached to the exception A failed task report, or an execution failure that already carries the current batch report

Examples:

  • A missing configuration returns 1 from ordinary validate and 2 from fast validate. A missing DBC follows the same distinction.
  • An explicitly configured but missing missing.asc is not read by static validate/explain, so those commands may return 0. The actual run reads it and returns 1. Static success does not prove that the log can be read.
  • An ordinary batch using jobs loads each referenced analysis configuration after its worker starts. If a child configuration references an empty directory or an unmatched glob, that task is marked failed, the batch returns 1, and a batch task report is available.
  • An ordinary batch using template/inputs expands the template and constructs each child plan while loading the batch configuration. For example, a valid template may have two CSV sources, primary and auxiliary. With source: primary, only primary is replaced by inputs. If the retained auxiliary source references an empty directory or an unmatched glob, loading returns 2: no task has started and no task report exists. This is not limited to expansion errors in batch inputs itself. See load_batch_plan for the loading contract.
  • fast batch checks its supported CAN child configurations before dispatch. Preparation errors such as empty directories or unmatched globs return 2, without starting tasks or producing task reports. It does not support the two-CSV template above.
  • If the output path is an existing ordinary file, ordinary run, ordinary batch run, and fast run return 1; fast batch returns 2. A file error while writing the final fast batch report can also return 2 after child tasks have succeeded. Their completed outputs remain available.

On a nonzero exit, retain stderr and inspect any task reports already written. A nonzero exit does not mean that all published results were rolled back. Failed or cancelled batches must not count unfinished tasks as successes.

See several configuration errors at once

Run chronaxis validate plan.yaml --format json. When fields can be parsed, it lists independently detectable errors together, such as invalid backend, batch_rows and table format. Each diagnostic retains available source, path, line, column, code, message and hint. JSON uses an errors array even for one error; the former singular error field is no longer returned. Text mode numbers the errors.

Repair malformed YAML/JSON syntax first. An invalid source or signal can prevent dependent checks; blocked_checks explains those checks instead of inventing values. Run validation again after repairs. A configuration error prevents run from starting analysis. Static checks may read DBC metadata when configuration defaults require it, but do not read business message contents.

The run summary returns the actual manifest path; batch run returns the actual report path. Output filenames receive a timestamp suffix; see result publication.

AI guide

chronaxis guide prints the absolute Chinese README path; chronaxis guide --language en selects English. Neither writes files. Optional --output new-file.zip saves a complete bilingual copy; its parent must exist and the target must be new. See AI guide.

Chronaxis user manual · Offline edition