Skip to content

Latest commit

 

History

History
95 lines (62 loc) · 6.17 KB

README.md

File metadata and controls

95 lines (62 loc) · 6.17 KB
Raw

Serialbox Python Tools

This folder contains the Serialbox python tools.

  • pp_ser.py - the Serialbox preproccessor to add the serialization directives to a Fortran program. For documentation see the docs folder.
  • serialbox - the Serialbox python module that allows de-serializing and visualizing data stored by Serialbox
  • test.py - the unittests for the Python Tools

Style Guide for Python Code in Serialbox

All the Python code in Serialbox should follow the PEP 8 coding convention. https://www.python.org/dev/peps/pep-0008/

The serialbox Python Module

The serialbox python module makes use of the C++ Serialbox wrapper to read serialized data written by Serialbox. For convenience the required dynamic library is distributed with an installation of Serialbox.

Installation

Build Serialbox as per the installation and building instructions. CMake installs the serialbox module to $CMAKE_INSTALL_PREFIX/python/serialbox with make install in the build directory.

To use the serialbox python module either

  • Add $CMAKE_INSTALL_PREFIX/python to your $PYTHONPATH
  • cd to $CMAKE_INSTALL_PREFIX/python before executing any python scripts
  • Or copy $CMAKE_INSTALL_PREFIX/python/serialbox to your desired python modulepath

Reading and Visualizing Serialized Data

In our example we want to look at serialized data from another process:

datapath = "/scratch/jenkins/data/double/normal/"

Reading

First, we load the Serializer from the serialbox module. We obtain the data by initializing the Serializer by giving it the path to the data:

from serialbox import Serializer
ser = Serializer(datapath)

By default the Serializer loads the data stored in Field.json. If your data is stored with a different prefix, e.g. Error.json, then you can give the different prefix, e.g. Error as an additional argument to the serializer:

err = Serializer(datapath, "Error")

In python the key-value format of the serializer is represented as a hash table. For example, if a savepoint is represented with the keys TimeIntegratorUnittest.DoRKStage-out, LargeTimeStep=0 and RKStageNumber=1 then this savepoint can be accessed as:

savepoint =  ser['TimeIntegratorUnittest.DoRKStage-out']['LargeTimeStep'][0]['RKStageNumber'][1]

To then load the pp variable from the savepoint simply add pp to the hash table and we get:

data = savepoint['pp']

data will now contain a numpy array that you can directly use with matplotlib and numpy. The interface allows removing the halo from the data by specifing "inner" as an additional argument:

data = savepoint['pp', 'inner']

The serializer allows dynamic exploration of the data. The representation of the Serializer object returns only the stored savepoints. For our dataset the result executed in a ipython shell will look like this:

In [4]: ser
Out[4]: { 'FastWavesSCUnittest.UV-out' = [...], 'AdvectionPDBottUnittest.Init-in' = [...], 'AdvectionPDBottUnittest.DoTracers-out' = [...], 'ConvertTemperatureUnittest.DoT-out' = [...], 'SedimentationUnittest.DoTracers-out' = [...], 'VerticalDiffusionUnittest.PrepareStep-in' = [...], 'VerticalDiffusionUnittest.DoUVWT-in' = [...], 'RelaxationUnittest.Apply-out' = [...], 'ConvertTemperatureUnittest.DoTP-in' = [...], 'FastWavesSCUnittest.WPPTP-out' = [...], 'VerticalDiffusionUnittest.DoTracers-in' = [...], 'RelaxationUnittest.Apply-in' = [...], 'ConvertTemperatureUnittest.DoTP-out' = [...], 'FastWavesSCUnittest.DoSmallStep-in' = [...], 'TimeIntegratorUnittest.DoRKStage-out' = [...], 'HorizontalDiffusionUnittest.DoStep-out' = [...], 'VerticalAdvectionUnittest.DoUVW-in' = [...], 'AdvectionPDBottUnittest.RecalculateDensity-in' = [...], 'HorizontalDiffusionUnittest.ColdPool-in' = [...], 'FastWavesSCUnittest.ExplicitDivergence-out' = [...], 'HorizontalAdvectionUnittest.DoWWCon-out' = [...], 'FastWavesSCUnittest.LHS-in' = [...], 'FastWavesSCUnittest.AllSteps-in' = [...], 'SaturationAdjustmentUnittest.Apply-in' = [...], 'CoriolisUnittest.Apply-in' = [...], 'TimeIntegratorUnittest.DoRKStage-in' = [...], 'TimeIntegratorUnittest.DoStep-out' = [...], 'FastWavesSCUnittest.RHS-in' = [...], 'CoriolisUnittest.Apply-out' = [...], 'ConvertTemperatureUnittest.DoT-in' = [...], 'VerticalAdvectionUnittest.DoPPTP-in' = [...], 'FastWavesSCUnittest.UV-in' = [...], 'HorizontalDiffusionUnittest.ColdPool-out' = [...], 'FastWavesSCUnittest.Init-out' = [...], 'VerticalAdvectionUnittest.DoPPTP-out' = [...], 'HorizontalAdvectionUnittest.DoUV-in' = [...], 'TimeIntegratorUnittest.DoStep-in' = [...], 'FastWavesSCUnittest.WPPTP-in' = [...], 'ConstantFields' = [...], 'DycoreUnittest.DoStep-out' = [...], 'FastWavesSCUnittest.ExplicitDivergence-in' = [...], 'FastWavesSCUnittest.DoSmallStep-out' = [...], 'AdvectionPDBottUnittest.Init-out' = [...], 'VerticalAdvectionUnittest.DoUVW-out' = [...], 'AdvectionPDBottUnittest.DoTracers-in' = [...], 'HorizontalAdvectionUnittest.DoPPTP-out' = [...], 'FastWavesSCUnittest.AllSteps-out' = [...], 'AdvectionPDBottUnittest.RecalculateDensity-out' = [...], 'HorizontalAdvectionUnittest.DoWWCon-in' = [...], 'VerticalDiffusionUnittest.DoUVWT-out' = [...], 'HorizontalAdvectionUnittest.DoUV-out' = [...], 'FastWavesSCUnittest.Init-in' = [...], 'VerticalDiffusionUnittest.DoTracers-out' = [...], 'VerticalDiffusionUnittest.PrepareStep-out' = [...], 'FastWavesSCUnittest.RHS-out' = [...], 'FastWavesSCUnittest.LHS-out' = [...], 'HorizontalAdvectionUnittest.DoPPTP-in' = [...], 'HorizontalDiffusionUnittest.DoStep-in' = [...], 'DycoreUnittest.DoStep-in' = [...], 'SedimentationUnittest.DoTracers-in' = [...], 'SaturationAdjustmentUnittest.Apply-out' = [...] }

Visualizing

To help visualizing the data serialbox contains a built-in Visualizer based on matplotlib. The visualizer expects two arguments:

  • The 3D numpy array loaded by the serializer
  • The name of the plot

In order to visualize the pp field stored in data from the code above it is sufficient to simply execute

from serialbox import Visualizer
Visualizer(data, 'pp')

This will open a window displaying the data:

Visualizer window

Each k level is represented as a simple slice in the window.