Begin tracking which lines of code are executed during a program run.

Indicate the start of a coverage measurement session.

When beginning a test suite, profiling run, or any code segment where execution coverage needs to be recorded.

Used in test scripts, performance profiling scripts, coverage reporting utilities, and interactive debugging sessions with coverage.py.

cov.start() where cov is an instance of coverage.Coverage

object.method() with object being a coverage.Coverage instance and method start()

cov: coverage.Coverage instance

• coverage.Coverage
• coverage.stop()
• coverage.report()
• coverage.xml_report()
• coverage.html_report()

• cov.start() vs cov.start(when='test') – specifying when to start coverage
• using coverage.erase() before start to clear previous data.

Calling cov.start() multiple times without intervening cov.stop() leads to overlapping coverage data; forgetting to call cov.stop() before generating a report results in incomplete data; importing coverage but failing to instantiate a Coverage object before calling start(); invoking cov.start() after the code under test has already executed, missing early lines; confusing coverage.start() with unittest.TestCase.start() which serves a different purpose.

coverage.stop() – halts coverage measurement; coverage.report() – outputs a textual coverage summary; sys.settrace() – provides lower‑level tracing without the coverage report features.

Coming from C++: may expect RAII‑style scope guards for coverage, but in Python you must explicitly call start() and stop(); coming from JavaScript: may assume automatic coverage collection, whereas Python requires explicit invocation.

• coverage.stop
• coverage.report
• coverage.html_report
• coverage.xml_report

Use coverage.start() only when you need detailed line‑by‑level execution data; enabling coverage adds measurable runtime overhead (often 2× slowdown); coverage does not trace native extensions or subprocesses unless specifically configured via the config file.

Think of coverage.start() as pressing the record button on a performance recorder before a play begins, ensuring every line spoken is captured for later review.

After calling cov.stop(), use cov.report() or cov.html_report() to generate human‑readable coverage reports.