Skip to content

Correct and improve log messages

Please develop a plan to correct and improve the log messages as follows:

  1. correct the skipped/executed counts in the final summary completed message in this example:
(py311) PS C:\dev\causaliq\causaliq-research> cqflow run papers/variable-ordering/workflows/var-ordering.yml --log-level all --mode run
2026-10-29 21:11:13 [causaliq-workflow] LOADING workflow from: papers\variable-ordering\workflows\var-ordering.yml
2026-10-29 21:11:13 [causaliq-workflow] VALIDATING workflow syntax and parameters...
2026-10-29 21:11:15 [causaliq-workflow] VALIDATED workflow successfully
2026-10-29 21:11:15 [causaliq-workflow] EXECUTING workflow in run mode...
2026-10-29 21:11:15 [./legacy.yml] CALLING      Migrate legacy results into CausalIQ legacy.db cache
2026-10-29 21:11:15 [migrate_trace] SKIPPED      Migrate legacy graphs for series BNLEARN/HC_STD [series=BNLEARN/HC_STD, network=asia, sample_size=1000]
2026-10-29 21:11:15 [migrate_trace] SKIPPED      Migrate legacy graphs for series BNLEARN/HC_OPT [series=BNLEARN/HC_OPT, network=asia, sample_size=1000]
2026-10-29 21:11:15 [migrate_trace] SKIPPED      Migrate legacy graphs for series BNLEARN/HC_BAD [series=BNLEARN/HC_BAD, network=asia, sample_size=1000]
2026-10-29 21:11:15 [./bnsl.yml] CALLING      Perform BNSL using CausalIQ algorithms putting results in learn.db
2026-10-29 21:11:15 [./bnsl_subflow.yml] CALLING      Run bnlearn hc with standard ordering
2026-10-29 21:11:15 [learn_graph] SKIPPED      Learn graph using bnlearn for BNLEARN/HC_STD [series=BNLEARN/HC_STD, network=asia, sample_size=1000]
2026-10-29 21:11:15 [./bnsl_subflow.yml] CALLING      Run bnlearn hc with optimal ordering
2026-10-29 21:11:15 [learn_graph] SKIPPED      Learn graph using bnlearn for BNLEARN/HC_OPT [series=BNLEARN/HC_OPT, network=asia, sample_size=1000]
2026-10-29 21:11:15 [./bnsl_subflow.yml] CALLING      Run bnlearn hc with worst ordering
2026-10-29 21:11:15 [learn_graph] SKIPPED      Learn graph using bnlearn for BNLEARN/HC_BAD [series=BNLEARN/HC_BAD, network=asia, sample_size=1000]
2026-10-29 21:11:15 [causaliq-workflow] COMPLETED 6 steps: 3 executed, 3 skipped (3 new entries)
  1. improve the error messages in the final ERRORS: summary message so that they include the matrix parameter values. If it is not already implemented, place a limit on the number of errors reported - perhaps 10, with a message such as "... and xxx more errors" detailing how many more errors there are.
(py311) PS C:\dev\causaliq\causaliq-research> cqflow run papers/variable-ordering/workflows/var-ordering.yml --log-level all --mode run
2026-10-29 21:26:03 [causaliq-workflow] LOADING workflow from: papers\variable-ordering\workflows\var-ordering.yml
2026-10-29 21:26:03 [causaliq-workflow] VALIDATING workflow syntax and parameters...
2026-10-29 21:26:05 [causaliq-workflow] VALIDATED workflow successfully
2026-10-29 21:26:05 [causaliq-workflow] EXECUTING workflow in run mode...
2026-10-29 21:26:05 [./legacy.yml] CALLING      Migrate legacy results into CausalIQ legacy.db cache
2026-10-29 21:26:05 [migrate_trace] SKIPPED      Migrate legacy graphs for series BNLEARN/HC_STD [series=BNLEARN/HC_STD, network=asia, sample_size=10]
2026-10-29 21:26:05 [migrate_trace] SKIPPED      Migrate legacy graphs for series BNLEARN/HC_OPT [series=BNLEARN/HC_OPT, network=asia, sample_size=10]
2026-10-29 21:26:05 [migrate_trace] SKIPPED      Migrate legacy graphs for series BNLEARN/HC_BAD [series=BNLEARN/HC_BAD, network=asia, sample_size=10]
2026-10-29 21:26:05 [./bnsl.yml] CALLING      Perform BNSL using CausalIQ algorithms putting results in learn.db
2026-10-29 21:26:05 [./bnsl_subflow.yml] CALLING      Run bnlearn hc with standard ordering
2026-10-29 21:26:06 [learn_graph] FAILED       Learn graph using bnlearn for BNLEARN/HC_STD [series=BNLEARN/HC_STD, network=asia, sample_size=10]
2026-10-29 21:26:06 [./bnsl_subflow.yml] CALLING      Run bnlearn hc with optimal ordering
2026-10-29 21:26:06 [learn_graph] FAILED       Learn graph using bnlearn for BNLEARN/HC_OPT [series=BNLEARN/HC_OPT, network=asia, sample_size=10]
2026-10-29 21:26:06 [./bnsl_subflow.yml] CALLING      Run bnlearn hc with worst ordering
2026-10-29 21:26:07 [learn_graph] FAILED       Learn graph using bnlearn for BNLEARN/HC_BAD [series=BNLEARN/HC_BAD, network=asia, sample_size=10]
2026-10-29 21:26:07 [causaliq-workflow] COMPLETED 6 steps: 3 skipped, 3 failed
2026-10-29 21:26:07 [causaliq-workflow] ERRORS:
  Run bnlearn hc with standard ordering: Learn graph using bnlearn for BNLEARN/HC_STD: failed
  Run bnlearn hc with optimal ordering: Learn graph using bnlearn for BNLEARN/HC_OPT: failed
  Run bnlearn hc with worst ordering: Learn graph using bnlearn for BNLEARN/HC_BAD: failed

  1. Each call to a sub-workflow should have a CALLING message as now, but additionally a COMPLETED message just before it returns control to the calling workflow. This should detail the cummulative numbers of steps skipped, failed and executed etc. which this workflow and any sub-workflows it calls have performed.

  2. Replace the values of the --log-level (all, summary, none) with an integer value that represents the maximum depth at which messages are reported (maybe rename the parameter as log-depth?), as follows

  3. 0: no message output - equivalent to none now
  4. 1: summary messages from the top-level workflow are reported only - equivalent to summary now for a standalone workflow. This would be the default value.
  5. 2: include log messages from the top-level steps - both actions it performs, and calls to sub-workflows but not any messages from within sub-workflows - equivalent to all for a standalone workflow, but omits messages from any sub-workflows
  6. 3: all messages from top-level and next level down sub-workflows but not from any deeper sub-workflows

  7. 3+: the pattern continues, including messages at greater depths

  8. Delineate the transitions into and out of sub-workflows more clearly visually - perhaps by indenting messages according to call depth

Further work which is out of scope for this task

These are deliberately outside this task:

  1. Values audit

  2. For discovery, knowledge and analysis actions, move input-like fields out of result into parameters.

  3. Check completeness and usefulness of each action’s metadata.
  4. Update action contract tests and API metadata documentation.

  5. This is the step that turns inputs: {} into the real effective-parameter record.

  6. Metadata catalogue and generated documentation

  7. Add a central catalogue of each action’s inputs, results and provenance expectations.

  8. Generate API documentation and CLI discovery from it.

  9. Add contract tests that every registered action is catalogued.

  10. history step parameter

  11. Support standard, a list of field paths, or all.

  12. Populate history for one-to-one aggregate steps.

  13. Add workflow/cache tests and user-guide documentation.

  14. Reporting capabilities

  15. Add filter support to the analysis plot action.

  16. Add categorical group-by support to summarise.

  17. Reuse the shared reporting view and update analysis/workflow docs and tests.

  18. merge_graphs many-to-one metadata

  19. Define action-specific reductions for n:1 merges.

  20. Generic history copying is not sufficient for many-to-one actions.

  21. Add tests and document the reduction rules.

  22. Hat 2 clean-up

  23. Remove the deprecated WorkflowCache.update_entry().

  24. Remove remaining UPDATE-pattern remnants and dead write paths.
  25. Consolidate duplicate metadata flatteners.

  26. No behavioural change.

  27. Smaller metadata backlog

  28. Expand {{series}}-style placeholders in workflow step names.

  29. Improve exported evaluate_graph metrics JSON readability with line feeds.
  30. Clarify metrics for failed/empty input graphs so they are not presented as evaluation failures.