mmfutils API¶
Contents:
MMF Utils¶
Small set of utilities: containers and interfaces.
This package provides some utilities that I tend to rely on during development. Presently it includes some convenience containers, plotting tools, and a patch for including zope.interface documentation in a notebook.
(Note: If this file does not render properly, try viewing it through nbviewer.org)
Documentation: http://mmfutils.readthedocs.org
Source: https://bitbucket.org/mforbes/mmfutils
Issues: https://bitbucket.org/mforbes/mmfutils/issues
Build Status:
Table of Contents
1 MMF Utils
1.1 Installing
2 Usage
2.1 Containers
2.1.1 Object
2.1.1.1 Object Example
2.1.2 Container
2.1.2.1 Container Examples
2.2 Contexts
2.3 Interfaces
2.3.1 Interface Documentation
2.4 Parallel
2.5 Performance
2.6 Plotting
2.6.1 Fast Filled Contour Plots
2.7 Angular Variables
2.8 Debugging
2.9 Mathematics
3 Developer Instructions
3.1 Releases
4 Change Log
4.1 REL: 0.4.7
Installing¶
This package can be installed from from the bitbucket project:
pip install hg+https://bitbucket.org/mforbes/mmfutils
Usage¶
Containers¶
Object¶
The Object
object provides a base class to satisfy the following
use-case.
Serialization and Deferred Initialization: Consider a problem where
a class is defined through a few parameters, but requires extensive
initialization before it can be properly used. An example is a numerical
simulation where one passes the number of grid points \(N\) and a
length \(L\), but the initialization must generate large grids for
efficient use later on. These grids should not be pickled when the
object is serialized: instead, they should be generated at the end of
initialization. By default, everything in __dict__
will be pickled,
leading to bloated pickles. The solution here is to split initialization
into two steps: __init__()
should initialize everything that is
picklable, then init()
should do any further initialization,
defining the grid points based on the values of the picklable
attributes. To do this, the semantics of the __init__()
method are
changed slightly here. Object.__init__()
registers all keys in
__dict__
as self.picklable_attributes
. These and only these
attributes will be pickled (through the provided __getstate__
and
__setstate__
methods).
The intended use is for subclasses to set and defined all attributes
that should be pickled in the __init__()
method, then call
Object.__init__(self)
. Any additional initialization can be done
after this call, or in the init()
method (see below) and attributes
defined after this point will be treated as temporary. Note, however,
that unpickling an object will not call __init__()
so any additional
initialization required should be included in the init()
method.
Deferred initialization via the ``init()`` method: The idea here is
to defer any expensive initialization – especially that which creates
large temporary data that should not be pickled – until later. This
method is automatically called at the end of Object.__init__()
and
after restoring a pickle. A further use-case is to allow one to change
many parameters, then reinitialize the object once with an explicit call
to init()
.
Object Example¶
__init__() called
init() called
State(L=1.0, N=256)
One feature is that a nice repr()
of the object is produced. Now
let’s do a calculation:
True
Here we demonstrate pickling. Note that the pickle is very small, and
when unpickled, init()
is called to re-establish s.x
and
s.k
.
169
init() called
Another use case applies when init()
is expensive. If \(x\) and
\(k\) were computed in __init__()
, then using properties to
change both \(N\) and \(L\) would trigger two updates. Here we
do the updates, then call init()
. Good practice is to call
init()
automatically before any serious calculation to ensure that
the object is brought up to date before the computation.
init() called
Finally, we demonstrate that Object
instances can be archived using
the persist
package:
__init__() called
init() called
State(L=2.0, N=64)
Container¶
The Container
object is a slight extension of Object
that
provides a simple container for storing data with attribute and
iterative access. These implement some of the Collections Abstract Base
Classes from the python standard
library.
The following containers are provided:
Container
: Bare-bones container extending theSized
,Iterable
, andContainer
abstract ase classes (ABCs) from the standardcontainers
library.ContainerList
: Extension that acts like a tuple/list satisfying theSequence
ABC from thecontainers
library (but not theMutableSequence
ABC. Although we allow setting and deleting items, we do not provide a way for insertion, which breaks this interface.)ContainerDict
: Extension that acts like a dict satisfying theMutableMapping
ABC from thecontainers
library.
These were designed with the following use cases in mind:
- Returning data from a function associating names with each data. The
resulting
ContainerList
will act like a tuple, but will support attribute access. Note that the order will be lexicographic. One could use a dictionary, but attribute access with tab completion is much nicer in an interactive session. Thecontainers.nametuple
generator could also be used, but this is somewhat more complicated (though might be faster). Also, named tuples are immutable - here we provide a mutable object that is picklable etc. The choice betweenContainerList
andContainerDict
will depend on subsequent usage. Containers can be converted from one type to another.
Container Examples¶
Container(a=1, b='Hi there', c=2)
(1, 'Hi there', 2)
Container(a=1, b='Ho there', c=2)
Container(a=1, b='Ho there', c=2)
[[ 1. 1. 1. ..., 1. 1. 1.]
[ 1. 1. 1. ..., 1. 1. 1.]
[ 1. 1. 1. ..., 1. 1. 1.]
...,
[ 1. 1. 1. ..., 1. 1. 1.]
[ 1. 1. 1. ..., 1. 1. 1.]
[ 1. 1. 1. ..., 1. 1. 1.]]
Container(a=1, b='Ho there', c=2)
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
<ipython-input-9-c6cad315ac19> in <module>()
2 c1 = pickle.loads(pickle.dumps(c))
3 print c1
----> 4 c1.large_temporary_array
AttributeError: 'Container' object has no attribute 'large_temporary_array'
Contexts¶
The mmfutils.contexts
module provides two useful contexts:
NoInterrupt
: This can be used to susspend KeyboardInterrupt
exceptions until they can be dealt with at a point that is convenient. A
typical use is when performing a series of calculations in a loop. By
placing the loop in a NoInterrupt
context, one can avoid an
interrupt from ruining a calculation:
Note: One can nest NoInterrupt
contexts so that outer loops are also
interrupted.
Interfaces¶
The interfaces module collects some useful zope.interface tools for checking interface requirements. Interfaces provide a convenient way of communicating to a programmer what needs to be done to used your code. This can then be checked in tests.
Here is a broken implementation. We muck up the arguments to add
:
BrokenMethodImplementation: The implementation of add violates its contract
because implementation requires too many arguments.
Now we get add
right, but forget to define value
. This is only
caught when we have an object since the attribute is supposed to be
defined in __init__()
:
BrokenImplementation: An object has failed to implement interface <InterfaceClass __main__.IAdder>
The value attribute was not provided.
Finally, a working instance:
True
Interface Documentation¶
We also monkeypatch zope.interface.documentation.asStructuredText()
to provide a mechanism for documentating interfaces in a notebook.
IAdder
Interface for objects that support addition.
Attributes:
value -- Current value of objectMethods:
add(other) -- Return self + other.
Parallel¶
The mmfutils.parallel
module provides some tools for launching and
connecting to IPython clusters. The parallel.Cluster
class
represents and controls a cluster. The cluster is specified by the
profile name, and can be started or stopped from this class:
Waiting for connection file: ~/.ipython/profile_default/security/ipcontroller-client.json
INFO:root:Starting cluster: ipcluster start --daemonize --quiet --profile=default --n=3
Waiting for connection file: ~/.ipython/profile_default/security/ipcontroller-client.json
INFO:root:waiting for 3 engines
INFO:root:0 of 3 running
INFO:root:3 of 3 running
INFO:root:Stopping cluster: ipcluster stop --profile=default
True
Waiting for connection file: ~/.ipython/profile_default/security/ipcontroller-client.json
If you only need a cluster for a single task, it can be managed with a context. Be sure to wait for the result to be computed before exiting the context and shutting down the cluster!
Waiting for connection file: ~/.ipython/profile_default/security/ipcontroller-client.json
INFO:root:Starting cluster: ipcluster start --daemonize --quiet --profile=default --n=3
Waiting for connection file: ~/.ipython/profile_default/security/ipcontroller-client.json
INFO:root:waiting for 3 engines
INFO:root:0 of 3 running
INFO:root:3 of 3 running
INFO:root:Stopping cluster: ipcluster stop --profile=default
Waiting for connection file: ~/.ipython/profile_default/security/ipcontroller-client.json
True
If you just need to connect to a running cluster, you can use
parallel.get_client()
.
Performance¶
The mmfutils.performance
module provides some tools for high
performance computing. Note: this module requires some additional
packages including
numexp,
pyfftw, and the mkl
package installed by anaconda. Some of these require building system
libraries (i.e. the FFTW). However, the
various components will not be imported by default.
Here is a brief description of the components:
mmfutils.performance.blas
: Provides an interface to a few of the scipy BLAS wrappers. Very incomplete (only things I currently need).mmfutils.performance.fft
: Provides an interface to the FFTW usingpyfftw
if it is available. Also enables the planning cache and setting threads so you can better control your performance.mmfutils.performance.numexpr
: Robustly imports numexpr and disabling the VML. (If you don’t do this carefully, it will crash your program so fast you won’t even get a traceback.)mmfutils.performance.threads
: Provides some hooks for setting the maximum number of threads in a bunch of places including the MKL, numexpr, and fftw.
Plotting¶
Several tools are provided in mmfutils.plot
:
Fast Filled Contour Plots¶
mmfutils.plot.imcontourf
is similar to matplotlib’s plt.contourf
function, but uses plt.imshow
which is much faster. This is useful
for animations and interactive work. It also supports my idea of saner
array-shape processing (i.e. if x
and y
have different shapes,
then it will match these to the shape of z
). Matplotlib now provies
plt.pcolourmesh
which is similar, but has the same interface issues.
CPU times: user 9.77 ms, sys: 50 µs, total: 9.82 ms
Wall time: 9.86 ms
CPU times: user 43.5 ms, sys: 1.19 ms, total: 44.6 ms
Wall time: 44.7 ms
CPU times: user 426 ms, sys: 34.1 ms, total: 460 ms
Wall time: 450 ms
CPU times: user 2.39 ms, sys: 346 µs, total: 2.73 ms
Wall time: 2.74 ms
<matplotlib.collections.QuadMesh at 0x1209acd10>
Angular Variables¶
A couple of tools are provided to visualize angular fields, such as the phase of a complex wavefunction.
(<matplotlib.contour.QuadContourSet at 0x11558d8d0>,
<matplotlib.contour.QuadContourSet at 0x115630b10>)
Debugging¶
A couple of debugging tools are provided. The most useful is the
debug
decorator which will store the local variables of a function
in a dictionary or in your global scope.
(1.0, 2.0, 2.8284271247461903, 1.0)
Mathematics¶
We include a few mathematical tools here too. In particular, numerical integration and differentiation. Check the API documentation for details.
Developer Instructions¶
If you are a developer of this package, there are a few things to be aware of.
If you modify the notebooks in
docs/notebooks
then you may need to regenerate some of the.rst
files and commit them so they appear on bitbucket. This is done automatically by thepre-commit
hook in.hgrc
if you include this in your.hg/hgrc
file with a line like:%include ../.hgrc
Security Warning: if you do this, be sure to inspect the .hgrc
file carefully to make sure that no one inserts malicious code.
This runs the following code:
[NbConvertApp] Converting notebook doc/README.ipynb to rst
[NbConvertApp] Support files will be in README_files/
[NbConvertApp] Making directory README_files
[NbConvertApp] Making directory README_files
[NbConvertApp] Writing 29492 bytes to README.rst
We also run a comprehensive set of tests, and the pre-commit hook will fail if any of these do not pass, or if we don’t have complete code coverage. This uses nosetests and flake8. To run individal tests do one of:
python setup.py nosetests
python setup.py flake8
python setup.py check
python setup.py test # This runs them all using a custom command defined in setup.py
Here is an example:
/data/apps/anaconda/envs/work/lib/python2.7/site-packages/setuptools-19.1.1-py2.7.egg/setuptools/dist.py:284: UserWarning: Normalizing '0.4.7dev' to '0.4.7.dev0' running test running nosetests running egg_info writing requirements to mmfutils.egg-info/requires.txt writing mmfutils.egg-info/PKG-INFO writing top-level names to mmfutils.egg-info/top_level.txt writing dependency_links to mmfutils.egg-info/dependency_links.txt reading manifest file 'mmfutils.egg-info/SOURCES.txt' writing manifest file 'mmfutils.egg-info/SOURCES.txt' nose.config: INFO: Set working dir to /Users/mforbes/work/mmfbb/mmfutils nose.config: INFO: Ignoring files matching ['^\.', '^_', '^setup\.py$'] nose.plugins.cover: INFO: Coverage report will include only packages: ['mmfutils'] INFO:root:Patching zope.interface.document.asStructuredText to format code INFO:root:Patching flake8 for issues 39 and 40 Doctest: mmfutils.containers.Container ... ok Doctest: mmfutils.containers.ContainerDict ... ok Doctest: mmfutils.containers.ContainerList ... ok Doctest: mmfutils.containers.Object ... ok Doctest: mmfutils.debugging.debug ... ok Doctest: mmfutils.debugging.persistent_locals ... ok Doctest: mmfutils.interface.describe_interface ... ok Doctest: mmfutils.math.differentiate.differentiate ... ok Doctest: mmfutils.math.differentiate.hessian ... ok Test the Richardson extrapolation for the correct scaling behaviour. ... ok Doctest: mmfutils.math.integrate.Richardson ... ok Doctest: mmfutils.math.integrate.exact_add ... ok Doctest: mmfutils.math.integrate.exact_sum ... ok Doctest: mmfutils.math.integrate.mquad ... /Users/mforbes/work/mmfbb/mmfutils/mmfutils/math/integrate/__init__.py:1: RuntimeWarning: divide by zero encountered in double_scalars """Integration Utilities. WARNING:root:mquad:MinStepSize: Minimum step size reached. (5.94368304574e-19 < 6.50521303491e-19) Singularity possible (err = 0.0). WARNING:root:mquad:MinStepSize: Minimum step size reached. (5.94368304574e-19 < 6.50521303491e-19) Singularity possible (err = 1.98122768191e-19). ok Doctest: mmfutils.math.integrate.quad ... ok Doctest: mmfutils.math.integrate.rsum ... ok Doctest: mmfutils.math.integrate.ssum_inline ... ok Doctest: mmfutils.math.integrate.ssum_python ... ok Test directional first derivatives ... ok Test directional second derivatives ... ok Doctest: mmfutils.optimize.bracket_monotonic ... ok Doctest: mmfutils.performance.fft.resample ... ok Doctest: mmfutils.performance.numexpr ... ok mmfutils.tests.test_containers.TestContainer.test_container_delattr ... ok Test persistent representation of object class ... ok Check that the order of attributes defined by ... ok mmfutils.tests.test_containers.TestContainerConversion.test_conversions ... ok mmfutils.tests.test_containers.TestContainerDict.test_container_del ... ok mmfutils.tests.test_containers.TestContainerDict.test_container_setitem ... ok mmfutils.tests.test_containers.TestContainerList.test_container_delitem ... ok mmfutils.tests.test_containers.TestObject.test_empty_object ... ok Test persistent representation of object class ... ok mmfutils.tests.test_containers.TestPersist.test_archive ... ok Doctest: mmfutils.tests.test_containers.Issue4 ... ok mmfutils.tests.test_debugging.TestCoverage.test_coverage_1 ... ok mmfutils.tests.test_debugging.TestCoverage.test_coverage_2 ... ok mmfutils.tests.test_debugging.TestCoverage.test_coverage_3 ... ok mmfutils.tests.test_debugging.TestCoverage.test_coverage_exception ... ok Test 3rd order differentiation ... ok mmfutils.tests.test_interface.TestInterfaces.test_verifyBrokenClass ... ok mmfutils.tests.test_interface.TestInterfaces.test_verifyBrokenObject1 ... ok mmfutils.tests.test_interface.TestInterfaces.test_verifyBrokenObject2 ... ok mmfutils.tests.test_interface.TestInterfaces.test_verifyClass ... ok mmfutils.tests.test_interface.TestInterfaces.test_verifyObject ... ok Doctest: mmfutils.tests.test_interface.Doctests ... ok mmfutils.tests.test_monkeypatchs.TestCoverage.test_cover_monkeypatchs ... INFO:root:Patching flake8 for issues 39 and 40 ok mmfutils.tests.test_monkeypatchs.TestCoverage.test_flake8_patch_err ... INFO:root:Patching flake8 for issues 39 and 40 ok [ProfileCreate] Generating default config file: u'/var/folders/m7/dnr91tjs4gn58_t3k8zp_g000000gn/T/tmp9itx0A/profile_testing/ipython_config.py' [ProfileCreate] Generating default config file: u'/var/folders/m7/dnr91tjs4gn58_t3k8zp_g000000gn/T/tmp9itx0A/profile_testing/ipython_kernel_config.py' [ProfileCreate] Generating default config file: u'/var/folders/m7/dnr91tjs4gn58_t3k8zp_g000000gn/T/tmp9itx0A/profile_testing/ipcontroller_config.py' [ProfileCreate] Generating default config file: u'/var/folders/m7/dnr91tjs4gn58_t3k8zp_g000000gn/T/tmp9itx0A/profile_testing/ipengine_config.py' [ProfileCreate] Generating default config file: u'/var/folders/m7/dnr91tjs4gn58_t3k8zp_g000000gn/T/tmp9itx0A/profile_testing/ipcluster_config.py' INFO:root:Starting cluster: ipcluster start --daemonize --quiet --profile=testing1 --n=7 --ipython-dir="/var/folders/m7/dnr91tjs4gn58_t3k8zp_g000000gn/T/tmp9itx0A" WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... INFO:root:waiting for 1 engines INFO:root:0 of 1 running INFO:root:7 of 1 running INFO:root:Starting cluster: ipcluster start --daemonize --quiet --profile=testing_pbs --n=3 --ipython-dir="/var/folders/m7/dnr91tjs4gn58_t3k8zp_g000000gn/T/tmp9itx0A" WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... WARNING:root:No ipcontroller-client.json, waiting... INFO:root:waiting for 1 engines INFO:root:0 of 1 running INFO:root:3 of 1 running Simple test connecting to a cluster. ... INFO:root:waiting for 1 engines INFO:root:7 of 1 running ok Test deleting of cluster objects ... ok Test that starting a running cluster does nothing. ... ok Test that the PBS_NODEFILE is used if defined ... INFO:root:waiting for 1 engines INFO:root:3 of 1 running INFO:root:waiting for 3 engines INFO:root:3 of 3 running INFO:root:Stopping cluster: ipcluster stop --profile=testing_pbs --ipython-dir="/var/folders/m7/dnr91tjs4gn58_t3k8zp_g000000gn/T/tmp9itx0A" 2016-01-05 12:16:55.566 [IPClusterStop] Stopping cluster [pid=17497] with [signal=2] ok Test timeout (coverage) ... ok mmfutils.tests.test_parallel.TestCluster.test_views ... DEBUG:root:Importing canning map ok INFO:root:Stopping cluster: ipcluster stop --profile=testing1 --ipython-dir="/var/folders/m7/dnr91tjs4gn58_t3k8zp_g000000gn/T/tmp9itx0A" 2016-01-05 12:16:56.330 [IPClusterStop] Stopping cluster [pid=17461] with [signal=2] mmfutils.tests.test_performance_blas.Test_BLAS.test_daxpy ... ok mmfutils.tests.test_performance_blas.Test_BLAS.test_ddot ... ok mmfutils.tests.test_performance_blas.Test_BLAS.test_dnorm ... ok mmfutils.tests.test_performance_blas.Test_BLAS.test_zaxpy ... ok mmfutils.tests.test_performance_blas.Test_BLAS.test_zdotc ... ok mmfutils.tests.test_performance_blas.Test_BLAS.test_znorm ... ok mmfutils.tests.test_performance_fft.Test_FFT.test_fft ... ok mmfutils.tests.test_performance_fft.Test_FFT.test_fftn ... ok mmfutils.tests.test_performance_fft.Test_FFT_pyfftw.test_fft ... ok mmfutils.tests.test_performance_fft.Test_FFT_pyfftw.test_fft_pyfftw ... ok mmfutils.tests.test_performance_fft.Test_FFT_pyfftw.test_fftn ... ok mmfutils.tests.test_performance_fft.Test_FFT_pyfftw.test_fftn_pyfftw ... ok mmfutils.tests.test_performance_fft.Test_FFT_pyfftw.test_get_fft_pyfftw ... ok mmfutils.tests.test_performance_fft.Test_FFT_pyfftw.test_get_fftn_pyfftw ... ok mmfutils.tests.test_performance_threads.TestThreads.test_hook_mkl ... ok mmfutils.tests.test_performance_threads.TestThreads.test_hooks_fft ... ok mmfutils.tests.test_performance_threads.TestThreads.test_hooks_numexpr ... ok mmfutils.tests.test_performance_threads.TestThreads.test_set_threads_fft ... ok mmfutils.tests.test_performance_threads.TestThreads.test_set_threads_mkl ... ok mmfutils.tests.test_performance_threads.TestThreads.test_set_threads_numexpr ... ok Name Stmts Miss Cover Missing ------------------------------------------------------------ mmfutils 1 0 100% mmfutils.containers 85 0 100% mmfutils.debugging 47 0 100% mmfutils.interface 70 0 100% mmfutils.math 0 0 100% mmfutils.math.differentiate 61 0 100% mmfutils.math.integrate 193 0 100% mmfutils.monkeypatches 14 0 100% mmfutils.optimize 13 0 100% mmfutils.parallel 124 2 98% 15-16 mmfutils.performance 0 0 100% mmfutils.performance.blas 58 0 100% mmfutils.performance.fft 61 0 100% mmfutils.performance.numexpr 10 0 100% mmfutils.performance.threads 10 0 100% ------------------------------------------------------------ TOTAL 747 2 99% ---------------------------------------------------------------------- Ran 73 tests in 19.302s OK
Complete code coverage information is provided in
build/_coverage/index.html
.
Coverage report: 99%
Hot-keys on this page
n s m x c change column sorting
Module | statements | missing | excluded | coverage |
---|---|---|---|---|
Total | 747 | 2 | 71 | 99% |
mmfutils | 1 | 0 | 0 | 100% |
mmfutils.containers | 85 | 0 | 0 | 100% |
mmfutils.debugging | 47 | 0 | 3 | 100% |
mmfutils.interface | 70 | 0 | 14 | 100% |
mmfutils.math | 0 | 0 | 0 | 100% |
mmfutils.math.differentiate | 61 | 0 | 0 | 100% |
mmfutils.math.integrate | 193 | 0 | 16 | 100% |
mmfutils.monkeypatches | 14 | 0 | 4 | 100% |
mmfutils.optimize | 13 | 0 | 0 | 100% |
mmfutils.parallel | 124 | 2 | 8 | 98% |
mmfutils.performance | 0 | 0 | 0 | 100% |
mmfutils.performance.blas | 58 | 0 | 6 | 100% |
mmfutils.performance.fft | 61 | 0 | 5 | 100% |
mmfutils.performance.numexpr | 10 | 0 | 7 | 100% |
mmfutils.performance.threads | 10 | 0 | 8 | 100% |
Releases¶
We try to keep the repository clean with the following properties:
- The default branch is stable: i.e. if someone runs
hg clone
, this will pull the latest stable release. - Each release has its own named branch so that e.g.
hg up 0.4.6
will get the right thing. Note: this should update to the development branch, not the default branch so that any work committed will not pollute the development branch (which would violate the previous point).
To do this, we advocate the following proceedure.
``hg up <version>``: Make sure this is the correct development branch, not the default branch. (Check by
hg up default
which should take you to the default branch.)Work: Do your work, committing as required with messages as shown in the repository with the following keys:
DOC
: Documentation changes.API
: Changes to the exising API. This could break old code.EHN
: Enhancement or new functionality. Without anAPI
tag, these should not break existing codes.BLD
: Build system changes (setup.py
,requirements.txt
etc.)TST
: Update tests, code coverage, etc.BUG
: Address an issue as filed on the issue tracker.BRN
: Start a new branch (see below).REL
: Release (see below).WIP
: Work in progress. Do not depend on these! They will be stripped. This is useful when testing things like the rendering of documentation on bitbucket etc. where you need to push an incomplete set of files. Please collapse and strip these eventually when you get things working.CHK
: Checkpoints. These should not be pushed to bitbucket!
``python setup.py test``: Make sure the tests pass. (
hg com
will do this automatically if you have linked the.hgrc
file as discussed above.Update Docs: Update the documentation if needed. To generate new documentation run:
cd doc sphinx-apidoc -eTE ../mmfutils -o source rm source/mmfutis.tests.*
Include any changes at the bottom of this file (
doc/README.ipynb
).Edit any new files created (titles often need to be added) and check that this looks good with
make html open build/html/index.html
Look especially for errors of the type
WARNING: document isn't included in any toctree
. This indicates that you probably need to add the module to an upper level.. toctree::
. Also look forWARNING: toctree contains reference to document u'...' that doesn't have a title: no link will be generated
. This indicates you need to add a title to a new file. For example, when I added themmf.math.optimize
module, I needed to update the following:
.. doc/source/mmfutils.rst
mmfutils
========
.. toctree::
...
mmfutils.optimize
...
.. doc/source/mmfutils.optimize.rst
mmfutils.optimize
=================
.. automodule:: mmfutils.optimize
:members:
:undoc-members:
:show-inheritance:
``hg histedit``: (or
hg rebase
, orhg strip
as needed) Clean up the repo before you push. Branches should generally be linear unless there is an exceptional reason to split development.Release: First edit
mmfutils/__init__.py
and update the version number by removing thedev
part of the version number. Commit only this change and then push only the branch you are working on:hg com -m "REL: <version>" hg push -b .
Create a pull request on the development fork from your branch to
default
on the release project bitbucket. Review it, fix anything, then accept the PR and close the branch.Start new branch: On the same development branch (not
default
), increase the version number inmmfutils/__init__.py
and adddev
: i.e.:__version__ = '0.4.7dev'
Then create this branch and commit this:
hg branch "0.4.7"
hg com -m "BRN: Started branch 0.4.7"
- Update MyPI index.
- Optional: Update any
setup.py
files that depend on your new features/fixes etc.
Change Log¶
REL: 0.4.7¶
API changes:
- Added
mmfutils.interface.describe_interface()
for inserting interfaces into documentation. - Added some DVR basis code to
mmfutils.math.bases
. - Added a diverging colormap and some support in
mmfutils.plot
. - Added a Wigner Ville distribution computation in
mmfutils.math.wigner
- Added
mmfutils.optimize.usolve
andubrentq
for finding roots with`uncertanties
<https://pythonhosted.org/uncertainties/>`__ support.
Issues:
- Resolve issue #8: Use
`ipyparallel
<https://github.com/ipython/ipyparallel>`__ now. - Resolve issue #9: Use pytest rather than
nose
(which is no longer supported). - Resolve issue #10: PYFFTW wrappers now support negative
axis
andaxes
arguments. - Address issue #11: Preliminary version of some DVR basis classes.
- Resolve issue #12: Added solvers with
`uncertanties
<https://pythonhosted.org/uncertainties/>`__ support.