wearpants/instrument

Python observability & instrumentation primitives for metrics and benchmarking

instrument's Issues

Rename to CsvDirectory

better installation instructions

explain 'batteries included' philosophy

document install targets/ requirements files

Add rest()

That does all()-first()

Think about how to optimize common first_and_rest() use case

Fix tmp directory in doctests for Travis

Because stupidness

Consider __getattr__ pass through

Maybe setattr too

remove dependency on numpy for table output

Use lists + stdlib statistics

better names?

maybe instument.reducer / instrument.producer / instrument.function

Use addCleanup() in tests

instead of teardown

document numpy

explaind how what the numpy module does, (what charts & stats mean, w/ some wikipedia links)
include some sample charts!

reducer optimization

would be helpful to have a version of measure_reduce that can use len() on first arg like measure_produce does on return value

Document using throw(stopiteration) with all() for early exit

modularize dependencies

Use setuptools extras: http://pythonhosted.org/setuptools/setuptools.html#declaring-extras-optional-features-with-their-own-dependencies

Pass through __length_hint__ on iterables

See pep

integrate sphinx doctests into nose/test runner

has no one done this already?!?!?

integrate sphinx doctests runner with traditional unit tests

what is sensible behavior for empty iterables?

Currently all/first/rest will ouput 0 items in 0.00 seconds - is this sensible, or should they just output nothing?

Maybe leave it up to metric functions to filter it out

Remove Metric from output names

CsvFile.metric looks nicer

add next steps / future plans doc

with notes about 0.5+

Pass through close(), send() & throw() to underlying iterable

this is dangerously close to supporting coroutines, but probably a good idea

tox support

add support for tox (multiple python versions)

Use __slots__ on inner utility classes

Whack with PEP 8

nose support

suport running tests under nose?

Add metric callbacks for garbage collection

so cool: https://docs.python.org/3/library/gc.html#gc.callbacks

Change description to "instrumentation"

tests for numpy metric

Mocks? Introspect internals? meh. Maybe just verify the correct files are written.

Document concurrency model

Threads mainly, also multi processing

This is all about outputs since the instrumentation is stack local

improvements to measure_reduce/measure_produce, etc.

maybe measure_sequence_producer is a better name than measure_produce?

would be helpful to have a version of measure_reduce that can use len() on first arg like measure_produce does on return value

Allow setting count within a block

Sometimes you don't know the count before a block, but rather calculate it during. Do something like

with instrument.block(name='foo') as block:
     # do stuff to produce some_number
     block.count = some_number

Yup

Convert readme to markdown and expand

document csv module

a sentence or two + usage example

CSVMetric should take configurable buffer size

as argument to class constructor

standalone backwards-incompatibilities doc

instead of in changelog

Rename .func() to .decorate()

Decorators can be used to wrap a generator function so that each call returns a measured iterable. This behavior is available via measure_iter.func() and so forth. The name is a bit confusing, and it'd be better to make it an active verb like measure_iter.decorate()

Maybe?

document how to run tests

b/c even I forgot

Allow leaving off () from decorators when empty

Steal from the best: https://github.com/hynek/attrs/blob/master/src/attr/_make.py#L230

clean up temp files from doctests

Tests for csv metric

set output file/dir to a tmp, check contents for correctness & cleanup

Logging metric

A simple metric function using stdlib's logging

travis ci

get on travis

add instrument.except_*()

like prometheus, I think the following:

• except_block
• except_iter
• except_func

that only measure on exceptions