| name | tools-dftracer-utils |
| description | dftracer-utils — the Python binding package for the trace-utils CLI (split/merge/compact/reader/event_count on .pfw/.pfw.gz traces). Install verification and its relationship to the dftracer/pydftracer package trio. Load this skill before any trace post-processing step (session_split_traces, reader, event_count, comparator). |
| metadata | {"type":"software"} |
What it provides
dftracer-utils is the PyPI package (pip show dftracer-utils →
Summary: Python binding for dftracer-utils) backing the trace-processing MCP
tools this project uses for ALL trace work — split, merge, reader,
event_count, comparator, aggregator, stats, reconstruct, replay,
organize, pgzip, tar (see the dftracer-trace-utils skill's "always use
the MCP tools, never raw gzip/python scripts" rule). It is a dependency of
dftracer itself (Required-by: dftracer, dftracer-agents, dftracer-analyzer
on the framework's own venv) — installing dftracer normally installs this
transitively.
Verification
python -c "import dftracer_utils"
or check the MCP tool wrapper directly succeeds on a real trace file rather
than trusting pip show alone.
Relationship to dftracer / pydftracer
These are three separate PyPI packages that together make up "the dftracer
ecosystem" for a session:
- dftracer — C core, tracing runtime, CMake build. See [[tools-dftracer]].
- pydftracer — Python annotation/decorator API (
dftracer.python). See [[tools-pydftracer]].
- dftracer-utils (this skill) — trace post-processing bindings (split/merge/read).
A session can have a working C core and trace-utils (so session_split_traces/
event_count/reader all work fine) while STILL missing pydftracer — these
are independent failure modes; verifying one does not verify the others. Check
all three explicitly when diagnosing a "dftracer partially works" report.
Session-local vs shared venv
Same rule as [[tools-dftracer]]: never repair the shared framework venv's
dftracer-utils in place as a side effect of a session — install a
session-local dftracer (which pulls this in transitively) instead.
