Release v0.4.0 · fedarko/wotplot

New features

Added the s1_name and s2_name parameters to viz_imshow() and viz_spy() (#17).
- You can use these parameters to adjust the sequence name shown on the x or y axis, respectively.

Updated the default algorithm for finding shared k-mers between two strings. Rather than manually iterating through the suffix arrays produced by pydivsufsort.divsufsort() to identify shared k-mers, we now instead use pydivsufsort.common_substrings() -- which accomplishes the same thing much faster.
- For example, creating a dot plot of two random 1 Mbp sequences (using k = 20) takes 26.52 seconds (max memory usage 563.09 MiB) using the old method and 2.16 seconds (max memory usage 617.53 MiB) using the new method.
- The new method does have higher memory usage for long sequences -- to the point where some test cases crash Jupyter Notebook with the new method, but succeed (albeit after taking a long time) with the old method. If desired, you can choose to use the old method ("suff-only") by passing suff_only=True to the DotPlotMatrix() constructor.
When using the old shared-k-mer-finding method ("suff-only") and creating a self dot plot (i.e. comparing a sequence with itself), wotplot will now detect if the sequences are identical and reuse the sequence's suffix array (rather than unnecessarily creating the same suffix array twice).
- Note that there are still other, more dramatic ways to speed up the creation of self dot plots; this is a relatively small improvement.

Breaking change: Removed the binary parameter of the DotPlotMatrix() constructor.
- Now, all matrices are "not binary" by default. If you want to visualize a matrix in a binary way, you can now set binary=True when calling viz_imshow() or viz_spy() (this is analogous to the force_binary parameter in wotplot 0.3.0). Matrix visualization defaults to visualizing the matrix in color.
- This change was motivated by benchmarking -- making the matrix "not binary" didn't seem to impact construction speed. It does impact visualization speed slightly, but the effects of this can be (entirely, I think?) offset by using binary=True during visualization.
- Sorry for the breaking change. I think this makes the interface a lot more natural to use.
Abstracted code from the benchmarking notebook to a separate file in the docs/ folder, and tried to tidy it up.
Restructured and tested the SciPy version checking code.
Additional tests.
Added pytest-mock as a development dependency.

Fixed a bug where, if every single cell in the dot plot matrix was a match cell, the resulting dot plot would be empty (#19).

Add better documentation to wotplot._make._get_row() and wotplot._make._make().
Various updates to the README, tutorial notebook, and benchmarking notebook.
The tutorial now includes a more complex example of creating a grid of dot plots. The example (shown below) creates a grid figure comparing all pairs of sequences with each other.