Testing in @sasjs/core

Overview

Tests are executed on a real SAS server using the SASjs CLI (sasjs test), not locally. Each test is a self-contained .sas file that is submitted to the server, and results are collected in the sasjsresults folder.

Running Tests

npm test        # runs: npx @sasjs/cli test -t server

The -t server flag selects the target (server type) from sasjs/sasjsconfig.json.

Test Structure

Test files live under tests/ in subfolders by platform applicability:

tests/base — run on all platforms (SAS 9 and Viya)
tests/sas9only — SAS 9 only (metadata server macros)
tests/viyaonly — Viya only
tests/serveronly — SASjs Server only
tests/x-platform — cross-platform (both SAS 9 and Viya)
tests/ddlonly — DDL-related tests

Naming convention: <macroname>.test.sas, with numbered variants (<macroname>.test.1.sas, .test.2.sas, ...) for multiple tests of the same macro. File names are lowercase, matching the lint rules.

Test Flow

Init: tests/testinit.sas runs before every test (configured in sasjsconfig.json under testConfig.initProgram). It sets up a unique app location (mcTestAppLoc), the compute context, calls %mp_init(), and enables debug options when _debug is set.
Test body: the test file itself runs. It should use %mp_assert() to record results into work.test_results:

%mp_assert(
  iftrue=(&syscc=0),
  desc=Checking for error condition,
  outds=work.test_results
)

Term: tests/testterm.sas runs after every test (testConfig.termProgram). It adds a final assertion that &syscc=0, then writes the results as JSON via %webout(OPEN) / %webout(OBJ,TEST_RESULTS) / %webout(CLOSE).

Results

After a test run, check the sasjsresults folder:

testResults.json / testResults.xml / testResults.csv — per-test PASS/FAIL with descriptions and comments
logs/<testname>.log — the full SAS log for each test; check here first when a test fails
coverage.lcov — coverage data

Writing Tests — Things to Know

Tests follow the same Doxygen header and lint standards as regular macros (@file, @brief, <h4> SAS Macros </h4> listing macros used).
Macro calls are not terminated with semicolons: use %mp_assert(...) not %mp_assert(...);.
Use %mp_assert(iftrue=(...), desc=..., outds=work.test_results) for every check — always append to work.test_results.
When comparing datasets after a round trip (eg through JSON), do not assert proc compare SYSINFO=0 directly — SYSINFO is a bitmask that includes attribute differences (length, format, label) which round trips legitimately change. Mask it to data-related bits only (64=missing obs in compare, 128=extra obs in compare, 4096=unequal values, 32768=obs count differs). Note SYSINFO is a read-only automatic macro variable, so store the masked value in a new variable.
Be careful with character data round trips: cats() and the plain $ informat strip leading blanks; use trim() and $char where leading blanks must be preserved.
After any change, run npx sasjs lint.
Do not edit generated copies under sasjsbuild/ — they are refreshed by the CI build.

3.2 KiB Raw Permalink Blame History