maskarray -- Efficient layout mask handling in Python

** As seen on Floating-Sparse-ghetti-monster Hackathon 2019 **

The python module maskarray enables the efficient representation of binary (on-off) process masks, as found in many EDA fields, specifically designed with automization of ASIC layout mask data in mind.

Main features are:

• Import of mask data from (multiple) same-size pixel data images (.png, .bmp)
• Import mask data from ndarray (2D or 3D)
• Optional image data transposition during import
• Efficient storage in row-based Run-Length-Encoding
• Tracking of spatial embedding information (resolution and origin)
• Indexing either spatially (float-indices) or ordinal (int-indices)
• Extraction of an arbitrary subset of data (along X-, Y-, Layer-dimensions) via the slicing interface of a_maskarray[xslice, yslice, layerslice]
• Iterate parametrically with specification of iterator size over subsets

  for row in a_maskarray(height=0.03):
     for window in a_maskarray(width=0.03):
        # window has dimensions of 0.03 x 0.03 reference units

• Save/Load the format to/from disk
• Hash the data for each stored layer
• The data is immutable by design and can be converted to a dense ndarray

Structure

docs/ -- specification via plantUML
maskarray/ -- implementation
/ - README, pip installer (setup.py)

Usage

NOTE: maskarray uses cython to compile some performance-critical stuff and link it in back to the python world. If you like to distribute the resulting binaries you want to compile the package with a generic set of compiler options to ensure it won't segfault on systems with a lesser set of processor extensions than you have compiled it for on your machine. In that case set the environment variable MASKARRAY_GENERIC=1 before compilation.

Installation

Simply run pip install . after checkout. The maskarray package will be available after installation. Alternatively if you feel your heart lives for GNU make after all you may also decide to hit make install

Local build

Just for testing pursposes during development consider make compile. This will trigger a complete build in this local directory and does not install the produced result into your package repository. The module then will only be available from within the root directory.

Build documentation

maskarray contains a class documentation driven by sphinx:numpy_doc. To rebuild the documentation run make doc in the main project directory. A HTML documentation will be generated in the doc/ directory.

Style checks

Consistency is important! Therefore remember to check the things you type fulfill at least the minimal standards, i.e. for that repository I chose PEP8 to be that kind of minimal standard.

NOTE: requires pycodestyle installed.