DEP: use scipy instead of opencv for dilation

however, it doesn't look like either dilation method supports
scipy sparse matrices. maybe this is different in newer scipy
vsns / with scipy sparse arrays? check

.github/workflows/main.yml

@@ -54,16 +54,7 @@ jobs:
         run: which pip
         shell: bash -el {0}
 
-      # use a specific opencv-python version to prevent wasting ~30 minutes
-      # building opencv; from https://stackoverflow.com/a/63669919
-      - name: Install wotplot and pip dependencies (Python = 3.6)
-        if: ${{ matrix.python-version == '3.6' }}
-        run: |
-          pip install opencv-python==4.3.0.38
-          pip install -e .[dev]
-        shell: bash -el {0}
-
-      - name: Install wotplot and pip dependencies (Python > 3.6)
+      - name: Install wotplot and pip dependencies
         if: ${{ matrix.python-version != '3.6' }}
         run: pip install -e .[dev]
         shell: bash -el {0}

README.md

@@ -48,19 +48,6 @@ I'll try to put this on PyPI / conda eventually.
 
 ### Installation notes
 
-#### Speeding up the installation of `opencv-python` in Python 3.6 environments
-
-If you're using Python 3.6, installing `opencv-python` through pip can take a
-long time (~30 minutes).
-See [this Stack Overflow thread](https://stackoverflow.com/q/63669752)
-for details.
-
-I've found that
-[running `pip install opencv-python==4.3.0.38`](https://stackoverflow.com/a/63669919)
-before you install wotplot fixes this problem. See
-[our GitHub Actions workflow](https://github.com/fedarko/wotplot/blob/ce702b63bf790c41d02b0493e3a7eebda6fcec70/.github/workflows/main.yml#L62-L63)
-for an example of this.
-
 ## Performance
 
 This code is not very optimized, so I don't recommend using it for sequences

docs/Tutorial.ipynb

@@ -21,7 +21,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "5afadb83",
+   "id": "e22e84ef",
    "metadata": {},
    "source": [
     "## 1.1. Define a small dataset"
@@ -41,7 +41,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "5fe7b9c1",
+   "id": "43e7f3eb",
    "metadata": {},
    "source": [
     "## 1.2. Create a dot plot matrix"
@@ -50,7 +50,7 @@
   {
    "cell_type": "code",
    "execution_count": 3,
-   "id": "e1b20e21",
+   "id": "26a75572",
    "metadata": {
     "scrolled": true
    },
@@ -61,7 +61,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "5ede7540",
+   "id": "568ef0c1",
    "metadata": {},
    "source": [
     "## 1.3. Inspect the dot plot matrix"
@@ -70,7 +70,7 @@
   {
    "cell_type": "code",
    "execution_count": 4,
-   "id": "1a527b76",
+   "id": "4856c174",
    "metadata": {},
    "outputs": [
     {
@@ -91,7 +91,7 @@
   {
    "cell_type": "code",
    "execution_count": 5,
-   "id": "8f6c5155",
+   "id": "fe685aad",
    "metadata": {},
    "outputs": [
     {
@@ -113,7 +113,7 @@
   {
    "cell_type": "code",
    "execution_count": 6,
-   "id": "0fc68844",
+   "id": "65955531",
    "metadata": {},
    "outputs": [
     {
@@ -130,7 +130,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "6b6915f2",
+   "id": "f145fcc4",
    "metadata": {},
    "source": [
     "The `DotPlotMatrix` object we just created, `m`, contains some extra information about the way it was created (e.g. the value of `k` we used). But the most interesting part of it is the actual matrix describing the dot plot! This matrix is stored in the `mat` attribute."
@@ -139,7 +139,7 @@
   {
    "cell_type": "code",
    "execution_count": 7,
-   "id": "08e06a06",
+   "id": "83a8f0f0",
    "metadata": {},
    "outputs": [
     {
@@ -160,7 +160,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "dc3e7eb9",
+   "id": "b6240658",
    "metadata": {},
    "source": [
     "### 1.3.1. Why is this matrix 16x16? Shouldn't it be 18x18?\n",
@@ -174,7 +174,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "993fe20e",
+   "id": "656b8e02",
    "metadata": {},
    "source": [
     "### 1.3.2. What's the deal with sparse matrices?\n",
@@ -187,7 +187,7 @@
   {
    "cell_type": "code",
    "execution_count": 8,
-   "id": "be896fb7",
+   "id": "119a4a76",
    "metadata": {},
    "outputs": [
     {
@@ -222,7 +222,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "9d243cb6",
+   "id": "bbe62927",
    "metadata": {},
    "source": [
     "### 1.3.3. Sidenote about the matrix type\n",
@@ -231,7 +231,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "0aa62098",
+   "id": "d228686f",
    "metadata": {},
    "source": [
     "## 1.4. Visualize the dot plot matrix"
@@ -274,7 +274,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 10,
+   "execution_count": 11,
    "id": "40a1ea34",
    "metadata": {},
    "outputs": [
@@ -284,7 +284,7 @@
        "Text(0.5, 1.0, 'Top-to-bottom')"
       ]
      },
-     "execution_count": 10,
+     "execution_count": 11,
      "metadata": {},
      "output_type": "execute_result"
     },
@@ -311,6 +311,14 @@
     "axLeft.set_title('Bottom-to-top', fontsize=18)\n",
     "axRight.set_title('Top-to-bottom', fontsize=18)"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "77cdb212",
+   "metadata": {},
+   "outputs": [],
+   "source": []
   }
  ],
  "metadata": {

setup.py

@@ -41,7 +41,6 @@
         "numpy",
         "scipy",
         "matplotlib",
-        "opencv-python",
     ],
     # you know tbh i'm not sure if you NEED to have already installed scipy
     # before installing wotplot. i remember that trying to pip install numpy

wotplot/_viz.py

@@ -1,4 +1,5 @@
-import cv2
+import scipy
+from scipy.ndimage import binary_dilation
 import numpy as np
 from math import ceil
 from matplotlib import pyplot
@@ -148,9 +149,11 @@ def viz_binary(m, num_dilation_iterations="auto", title=None, ax=None,
             num_dilation_iterations = 4 + ceil((maxlen - 2500) / 2500)
 
     if num_dilation_iterations > 0:
-        matrix_to_show = cv2.dilate(
-            m.mat, DILATION_KERNEL, iterations=num_dilation_iterations
-        )
+        # scipy's dilation function doesn't seem to support sparse matrices...
+        # TODO TODO
+        matrix_to_show = scipy.sparse.coo_matrix(binary_dilation(
+            m.mat.toarray(), iterations=num_dilation_iterations
+        ).astype(m.mat.dtype))
     else:
         matrix_to_show = m.mat