Output & scripting¶

tx is built to be piped. Two rules make that work:

1. Data goes to stdout, diagnostics go to stderr. You can pipe any command's output to jq without ever receiving a warning.
2. Every command that prints a table also prints JSON or CSV via --output-format json|csv. Model-shaped output can also be emitted as tmdl or tmsl/bim.

Errors follow the same contract: --error-format json emits a single JSON error envelope on stderr (see the error codes reference).

Paths are the pipeline currency¶

ls --paths-only and find --paths-only emit one object path per line, and most commands accept paths as input:

# Format every object whose expression mentions CALCULATE
tx find "CALCULATE" --in expressions --paths-only | xargs -I{} tx format -p "{}"

# Count columns per table
tx ls --type column --output-format json |
  jq 'group_by(.path | split("/")[0]) | map({(.[0].path | split("/")[0]): length}) | add'

Querying live models¶

query runs DAX or DMV against a live model, with DAX Studio-style performance options: --trace (formula- vs storage-engine timings), --plan (logical and physical query plans), --cold (clear the cache first), and --runs N (benchmark with Avg/Min/Max/StdDev). Timings and plans print to stderr, so the rowset on stdout stays pipeable; with --output-format json they are folded into the result document instead.

# Server timings + query plan for a measure (needs workspace/server admin)
tx query -q 'EVALUATE ROW("Sales", [Total Sales])' --trace --plan

# Benchmark a heavy query cold, five runs
tx query --file heavy.dax --cold --runs 5

The trace-based options require admin rights on the endpoint; when unavailable they warn and the query still returns its rows.

Exit codes¶

diff is the exception by design: 0 = identical, 1 = differences found, 2 = error — so it drops straight into a CI gate.

CI¶

Flags that matter in pipelines:

• --non-interactive — disable all prompts; fail with an actionable error instead of hanging.
• --quiet — suppress spinners, progress, and hints.
• --ci github / --ci vsts on validate, test, and deploy — emit GitHub Actions / Azure DevOps logging commands to stderr, so findings annotate the PR.
• --trx <path> on validate, test, and bpa run — write results as a VSTEST .trx file.
• Color is stripped automatically when output is not a TTY, and the NO_COLOR convention is honored.

# GitHub Actions: lint and validate a model on every PR
- run: tx bpa run -m ./model.tmdl --non-interactive
- run: tx validate -m ./model.tmdl --ci github --non-interactive

DAX regression tests as a PR gate¶

tx test runs .dax queries with recorded expected results (details) and exits 1 on any drift. Because tests execute on a deployed model, the PR gate is: deploy the PR's model to a dev workspace, refresh it, then run the tests against it. On Azure DevOps with a service principal:

# Azure DevOps: deploy the PR model to the dev workspace, then gate on tests
- script: >
    tx test ./MyModel.SemanticModel/tests
    -s "$(DevWorkspace)" -d MyModel
    --auth spn --non-interactive
    --trx $(Agent.TempDirectory)/dax-tests.trx --ci vsts
  displayName: DAX regression tests
- task: PublishTestResults@2
  condition: always()
  inputs:
    testResultsFormat: VSTest
    testResultsFiles: $(Agent.TempDirectory)/dax-tests.trx

A non-zero exit fails the stage; --ci vsts additionally annotates each failing test. Accept intentional result changes by re-running tx test --update locally and committing the snapshot diff.

Ready-made workflow examples live in the samples folder.