pandalone: process data-trees with relocatable-paths

pandalone is a collection of utilities for working with hierarchical-data using relocatable-paths.

Release

0.5.0

Date

2020-05-14 23:59

Documentation

https://pandalone.readthedocs.org/

Source

https://github.com/pandalone/pandalone

PyPI repo

https://pypi.python.org/pypi/pandalone

Keywords

calculation, data, dependencies, engineering, excel, library, numpy, pandas, processing, python, resolution, scientific, simulink, tree, utility

Copyright

2015 European Commission (JRC-IET)

License

EUPL 1.1+

Currently only 2 portions of the envisioned functionality are ready for use:

• pandalone.xleash: A mini-language for "throwing the rope" around rectangular areas of Excel-sheets.
• pandalone.mappings: Hierarchical string-like objects that may be used for indexing, facilitating renaming keys and column-names at a later stage.

Our goal is to facilitate the composition of engineering-models from loosely-coupled components. Initially envisioned as an indirection-framework around pandas coupled with a dependency-resolver, every such model should auto-adapt and process only values available, and allow remapping of the paths accessing them, to run on renamed/relocated value-trees without component-code modifications.

It is an open source library written and tested on Python-3.6+ , Windows and Linux.

Note

The project, as of May-2015, is considered at an alpha-stage, without any released version in pypi yet.

Introduction

Overview

At the most fundamental level, an "execution" or a "run" of any data-processing can be thought like that:

.--------------.     _____________        .-------------.

; DataTree ; | | ; DataTree ;

;--------------; ==> | <cfunc_1> | ==> ;--------------;

; /some/data ; | <cfunc_2> | ; /some/data ;

; /some/other ; | ... | ; /some/other ;

; /foo/bar ; ; /foo/bar ;

'--------------' '--------------.

• The data-tree might come from json, hdf5, excel-workbooks, or plain dictionaries and lists. Its values are strings and numbers, numpy-lists, pandas or xray-datasets, etc.

• The component-functions must abide to the following simple signature:

  cfunc_do_something(pandelone, datatree)

  and must not return any value, just read and write into the data-tree.

• Here is a simple component-function:

  def cfunc_standardize(pandelone, datatree):
      pin, pon = pandelone.paths(),
      df = datatree.get(pin.A)
      df[pon.A.B_std] = df[pin.A.B] / df[pin.A.B].std()

• Notice the use of the relocatable-paths marked specifically as input or output.
• TODO: continue rough example in tutorial...

Quick-start

The program runs on Python-3.5+ and requires numpy, pandas and (optionally) win32 libraries along with their native backends.

pip install pandalone                 ## Use `--pre` if version-string has a build-suffix.

... but probably you need the following for xleash to work:

pip install pandalone[xlrd]

All "extras" are: test, doc, excel, pandas, xlrd, dev, all

In case you need the very latest from master branch :

pip install git+https://github.com/pandalone/pandalone.git

Or in to install in develop mode, with all dependencies needed for development, and with pre-commit hook for auto-formatting python-code with black, clone locally this project from the remote repo, and run:

pip install -e <pandalone-dr>[dev]
pre-commit install

Project files and folders

The files and folders of the project are listed below:

+--pandalone/       ## (package) Python-code
+--tests/           ## (package) Test-cases
+--doc/             ## Documentation folder
+--setup.py         ## (script) The entry point for `setuptools`, installing, testing, etc
+--requirements/    ## (txt-files) Various pip and conda dependencies.
+--README.rst
+--CHANGES.rst
+--AUTHORS.rst
+--CONTRIBUTING.rst
+--LICENSE.txt

Usage

Currently 2 portions of this library are ready for use: pandalone.xleash and pandalone.mappings

GUI usage

Attention

Desktop UI requires Python 3!

For a quick-'n-dirty method to explore the structure of the data-tree and run an experiment, just run:

$ pandalone gui

Excel usage

Attention

Excel-integration requires Python-3 and Windows or OS X!

In Windows and OS X you may utilize the excellent xlwings library to use Excel files for providing input and output to the experiment.

To create the necessary template-files in your current-directory you should enter:

$ pandalone excel

You could type instead pandalone excel {file_path} to specify a different destination path.

[TBD]

Python usage

Example python REPL (Read-Eval-Print Loop) example-commands are given below that setup and run an experiment.

First run python or ipython and try to import the project to check its version:

>>> import pandalone

>>> pandalone.__version__ ## Check version once more. '0.5.0'

>>> pandalone.__file__ ## To check where it was installed. # doctest: +SKIP /usr/local/lib/site-package/pandalone-...

If everything works, create the data-tree to hold the input-data (strings and numbers). You assemble data-tree by the use of:

• sequences,
• dictionaries,
• pandas.DataFrame,
• pandas.Series, and
• URI-references to other data-trees.

[TBD]

Getting Involved

This project is hosted in github. To provide feedback about bugs and errors or questions and requests for enhancements, use github's Issue-tracker.

Sources & Dependencies

To get involved with development, you need a POSIX environment to fully build it (Linux, OSX or Cygwin on Windows).

First you need to download the latest sources:

$ git clone https://github.com/pandalone/pandalone.git pandalone.git
$ cd pandalone.git

Virtualenv

You may choose to work in a virtualenv (isolated Python environment)_, to install dependency libraries isolated from system's ones, and/or without admin-rights (this is recommended for Linux/Mac OS).

Attention

If you decide to reuse stystem-installed packages using option --system-site-packages with virtualenv <= 1.11.6 (to avoid, for instance, having to reinstall numpy and pandas that require native-libraries) you may be bitten by bug #461 which prevents you from upgrading any of the pre-installed packages with pip.

Liclipse IDE

Within the sources there are two sample files for the comprehensive LiClipse IDE:

• eclipse.project
• eclipse.pydevproject

Remove the eclipse prefix, (but leave the dot(.)) and import it as "existing project" from Eclipse's File menu.

Another issue is caused due to the fact that LiClipse contains its own implementation of Git, EGit, which badly interacts with unix symbolic-links, such as the docs/docs, and it detects working-directory changes even after a fresh checkout. To workaround this, Right-click on the above file Properties --> Team --> Advanced --> Assume Unchanged

Then you can install all project's dependencies in `development mode using the setup.py script:

$ python setup.py --help                           ## Get help for this script.
Common commands: (see '--help-commands' for more)

  setup.py build      will build the package underneath 'build/'
  setup.py install    will install the package

Global options:
  --verbose (-v)      run verbosely (default)
  --quiet (-q)        run quietly (turns verbosity off)
  --dry-run (-n)      don't actually do anything
...

$ python setup.py develop                           ## Also installs dependencies into project's folder.
$ python setup.py build                             ## Check that the project indeed builds ok.

You should now run the test-cases to check that the sources are in good shape:

$ python setup.py test

Note

The above commands installed the dependencies inside the project folder and for the virtual-environment. That is why all build and testing actions have to go through python setup.py {some_cmd}.

If you are dealing with installation problems and/or you want to permantly install dependant packages, you have to deactivate the virtual-environment and start installing them into your base python environment:

$ deactivate
$ python setup.py develop

or even try the more permanent installation-mode:

$ python setup.py install                # May require admin-rights

Design

See architecture live-document.

FAQ

Why another XXX? What about YYY?

These are the knowingly related python projects:

• OpenMDAO: It has influenced pandalone's design. It is planned to interoperate by converting to and from it's data-types. But it is Python-2 only and its architecture needs attending from programmers (no setup.py, no official test-cases).
• PyDSTool: It does not overlap, since it does not cover IO and dependencies of data. Also planned to interoperate with it (as soon as we have a better grasp of it :-). It has some issues with the documentation, but they are working on it.
• xray: Pandas for higher dimensions; data-trees should in principle work with "xray".
• Blaze: NumPy and Pandas interface to Big Data; data-trees should in principle work with "blaze".
• netCDF4: Hierarchical file-data-format similar to hdf5; a data-tree may derive in principle from "netCDF4 ".
• hdf5: Hierarchical file-data-format, supported natively by pandas; a data-tree may derive in principle from "netCDF4 ".

Which other projects/ideas have you reviewed when building this library?

• bubbles ETL: Processing-pipelines for (mostly) categorical data.
• Data-protocols:
  • JTSKit, A utility library for working with JSON Table Schema in Python.
  • Data Packages
• Celery: Execute distributed asynchronous tasks using message passing on a single or more worker servers using multiprocessing, Eventlet, or gevent.
• Fuzzywuzzy and Jellyfish: Fuzzy string matching in python. Use it for writting code that can read coarsely-known column-names.
• "Other's people's messy data (and how not to hate it)", PyCon 2015(Canada) presentation by Mali Akmanalp.

Glossary

data-tree

The container of data consumed and produced by a :term`model`, which may contain also the model. Its values are accessed using path s. It is implemented by pandalone.pandata.Pandel as a mergeable stack of JSON-schema abiding trees of strings and numbers, formed with:

• sequences,
• dictionaries,
• pandas instances, and
• URI-references.

value-tree

That part of the data-tree that relates only to the I/O data processed.

model

A collection of component s and accompanying mappings.

component

Encapsulates a data-transformation function, using path to refer to its inputs/outputs within the value-tree.

path

A /file/like string functioning as the id of data-values in the data-tree. It is composed of step, and it follows the syntax of the JSON-pointer.

step pstep path-step The parts between between two conjecutive slashes(/) within a path. The Pstep facilitates their manipulation.

pmod pmods pmods-hierarchy mapping mappings Specifies a transformation of an "origin" path to a "destination" one (also called as "from" and "to" paths). The mapping always transforms the final path-step, and it can either rename or relocate that step, like that:

ORIGIN          DESTINATION   RESULT_PATH
------          -----------   -----------
/rename/path    foo       --> /rename/foo        ## renaming
/relocate/path  foo/bar   --> /relocate/foo/bar  ## relocation
/root           a/b/c     --> /a/b/c             ## Relocates all /root sub-paths.

The hierarchy is formed by Pmod instances, which are build when parsing the mappings list, above.

JSON-schema

The JSON schema is an IETF draft that provides a contract for what JSON-data is required for a given application and how to interact with it. JSON Schema is intended to define validation, documentation, hyperlink navigation, and interaction control of JSON data. You can learn more about it from this excellent guide, and experiment with this on-line validator.

JSON-pointer

JSON Pointer(6901) defines a string syntax for identifying a specific value within a JavaScript Object Notation (JSON) document. It aims to serve the same purpose as XPath from the XML world, but it is much simpler.