test

PnetCDF-python unit tests

This directory contains unit test programs in python. Detailed description of each test and run instructions are provided at the beginning of each file. All tests are expected to run with multiple MPI processes.

---

Running individual test programs

Make sure PnetCDF-python and all its dependent packages are installed first. Please refer to ../README. in the top directory for installation information.

• Use mpiexec command to run individual test programs. For example, command line below run tst_file_mode.py on 4 MPI processes.

  mpiexec -n 4 python tst_file_mode.py [file_output_dir]

• The optional argument file_output_dir enables the testing program to save the generated output files in the specified directory. Default is the current directory.

---

Overview of Test Programs

• tst_file

  • This series of test programs is focused on file creation and access through the File constructor, particularly with respect to the following aspects:
    • different access modes ("r+", "w", etc)
    • clobber option

• tst_dims

  • This series of tests is focused on dimension initialization, dimension methods, and their interactions with netCDF variables using the File object API. Particularly, these test program tests the following:
    • Dimension object basic attributes and methods including name, length
    • interactions with netCDF variable
      • different syntax for referencing associated dimensions at defining variables step
      • unlimited dimension length changes after adding/removing variable data

• tst_atts

  • This series of tests is focused on manipulating attributes using the File object API (for global attributes) and the Variable object API (for variable attributes):
    • define attributes of various data types with explicit methods or python-dictionary style syntax
    • attribute-based methods

• tst_var

  • This series of test programs writes data to or reads from variables within a netCDF file with different syntaxes and different access patterns using the Variable object interface. For data mode operations, both independent I/O and collective I/O are tested by default.

  • tst_var_indexer

    • Reading from or writing data to netCDF variable using slicing or indexer (numpy-style) syntax

  • tst_var_type

    • Writing data of heterogeneous data types to the defined variable

  • tst_var_put

    • This series of tests look into the process of writing data to a netCDF variable using explicit function-call style method concerning different needs of access patterns. Usually, each process is configured to write to a designated area within the netCDF variable.

  • tst_var_get

    • This series of tests is focused on reading data from a netCDF variable using explicit function-call style method with respect to different needs of access patterns. Usually, each process is configured to read from a designated area within the netCDF variable.

  • tst_var_iget/iput

    • This series of tests is focused on the non-blocking mode of variable operations mentioned above. The program usually posts read(iget) or write(iput) requests to access a netCDF variable using explicit function-call style method and calls the wait function to commit them.

  • tst_var_bput

    • This series of tests is focused on the buffered non-blocking mode of variable operations mentioned above. The program usually attaches a write buffer to the netCDF file, posts read(iget) or write(iput) requests to access a netCDF variable and then calls the wait function to commit.

• tst_default_format.py

  • Test set_default_format function for creating a number of netCDF files with default format.

• tst_rename.py

  • Test renameVariable/dim functions of a Dataset instance for renaming previously defined variables and dimensions.

• tst_version.py

  • Test version string of PnetCDF-Python.

• tst_wait.py

  • Test non-blocking APIs and then use wait/wait_all method of File class to flush out the pending I/O requests.

• tst_copy_attr.py

  • Copying an attribute from one file to another in python can be done without ncmpi_copy_att(). For exampl, this can be done in two lines of python codes below:

    att = source_file.get_att("history")
    destition_file.put_att('history', att)