automated tagging

Author: MarcoGorelli

.pre-commit-config.yaml

@@ -35,3 +35,12 @@ repos:
       language: python
       name: Check all notebooks appear in table of contents
       types: [jupyter]
+    - id: add-tags
+      entry: python scripts/add_tags.py
+      language: python
+      name: Add PyMC3 classes used to tags
+      types: [jupyter]
+      additional_dependencies:
+         - nbqa==1.1.1
+         - beautifulsoup4==4.9.3
+         - myst_parser==0.13.7

examples/case_studies/multilevel_modeling.ipynb

@@ -7,7 +7,7 @@
     "# A Primer on Bayesian Methods for Multilevel Modeling\n",
     "\n",
     ":::{post} 30 Aug, 2021\n",
-    ":tags: hierarchical\n",
+    ":tags: hierarchical, pymc3.Data, pymc3.Deterministic, pymc3.Exponential, pymc3.LKJCholeskyCov, pymc3.Model, pymc3.MvNormal, pymc3.Normal\n",
     ":category: intermediate\n",
     ":::"
    ]

examples/case_studies/rugby_analytics.ipynb

@@ -7,7 +7,7 @@
     "# A Hierarchical model for Rugby prediction\n",
     "\n",
     ":::{post} 30 Aug, 2021\n",
-    ":tags: hierarchical, sports\n",
+    ":tags: hierarchical, pymc3.Data, pymc3.Deterministic, pymc3.HalfNormal, pymc3.Model, pymc3.Normal, pymc3.Poisson, sports\n",
     ":category: intermediate\n",
     ":::"
    ]

examples/getting_started.ipynb

@@ -11,7 +11,7 @@
     "Note: This text is based on the [PeerJ CS publication on PyMC3](https://peerj.com/articles/cs-55/).\n",
     "\n",
     ":::{post} 30 Aug, 2021\n",
-    ":tags: glm, mcmc, exploratory analysis\n",
+    ":tags: exploratory analysis, glm, mcmc, pymc3.Data, pymc3.Deterministic, pymc3.DiscreteUniform, pymc3.Exponential, pymc3.GaussianRandomWalk, pymc3.HalfNormal, pymc3.Model, pymc3.Normal, pymc3.Poisson, pymc3.Slice, pymc3.StudentT\n",
     ":category: beginner\n",
     ":::"
    ]

scripts/__init__.py

@@ -0,0 +1,87 @@
+"""
+Automatically add tags to notebook based on which PyMC3 classes are used.
+
+E.g. if a notebook contains a section like
+
+    :::{post} 30 Aug, 2021
+    :tags: glm, mcmc, exploratory analysis
+    :category: beginner
+    :::
+
+in a markdown cell, and uses the class pymc3.Categorical, then this script
+will change that part of the markdown cell to:
+
+    :::{post} 30 Aug, 2021
+    :tags: glm, mcmc, exploratory analysis, pymc3.Categorical
+    :category: beginner
+    :::
+"""
+import sys
+from myst_parser.main import to_tokens, MdParserConfig
+import subprocess
+import os
+import json
+import argparse
+
+
+def main(argv=None):
+    parser = argparse.ArgumentParser()
+    parser.add_argument("files", nargs="*")
+    args = parser.parse_args(argv)
+
+    for file in args.files:
+        # Find which PyMC3 classes are used in the code.
+        output = subprocess.run(
+            [
+                "nbqa",
+                "scripts.find_pm_classes",
+                file,
+            ],
+            stdout=subprocess.PIPE,
+            text=True,
+        )
+        classes = {f"pymc3.{obj}" for obj in output.stdout.splitlines()}
+
+        # Tokenize the notebook's markdown cells.
+        with open(file, encoding="utf-8") as fd:
+            content = fd.read()
+        notebook = json.loads(content)
+        markdown_cells = "\n".join(
+            [
+                "\n".join(cell["source"])
+                for cell in notebook["cells"]
+                if cell["cell_type"] == "markdown"
+            ]
+        )
+        config = MdParserConfig(enable_extensions=["dollarmath", "colon_fence"])
+        tokens = to_tokens(markdown_cells, config=config)
+
+        # Find a ```{post} or :::{post} code block, and look for a line
+        # starting with tags: or :tags:.
+        tags = None
+        for token in tokens:
+            if token.tag == "code" and token.info.startswith("{post}"):
+                for line in token.content.splitlines():
+                    if line.startswith("tags: "):
+                        line_start = "tags: "
+                        original_line = line
+                        tags = {tag.strip() for tag in line[len(line_start) :].split(",")}
+                        break
+                    elif line.startswith(":tags: "):
+                        line_start = ":tags: "
+                        original_line = line
+                        tags = {tag.strip() for tag in line[len(line_start) :].split(",")}
+                        break
+
+        # If tags were found, then append any PyMC3 classes which might have
+        # been missed.
+        if tags is not None:
+            new_tags = ", ".join(sorted(tags.union(classes)))
+            new_line = f"{line_start}{new_tags}"
+            content = content.replace(original_line, new_line)
+            with open(file, "w", encoding="utf-8") as fd:
+                fd.write(content)
+
+
+if __name__ == "__main__":
+    exit(main())

scripts/add_tags.py

@@ -0,0 +1,57 @@
+"""
+Find all PyMC3 classes used in script.
+
+This'll find both call of
+
+    pymc3.Categorical(...
+
+and
+    
+    from pymc3 import Categorical
+    Categorical
+"""
+import ast
+import sys
+
+
+class ImportVisitor(ast.NodeVisitor):
+    def __init__(self, file):
+        self.imports = set()
+
+    def visit_ImportFrom(self, node: ast.ImportFrom):
+        if node.module.split(".")[0] == "pymc3":
+            for name in node.names:
+                if name.name[0].isupper():
+                    self.imports.add(name.name)
+
+
+class CallVisitor(ast.NodeVisitor):
+    def __init__(self, file, imports):
+        self.file = file
+        self.imports = imports
+        self.classes_used = set()
+
+    def visit_Call(self, node: ast.Call):
+        if isinstance(node.func, ast.Attribute):
+            if isinstance(node.func.value, ast.Name):
+                if node.func.value.id in {"pm", "pymc3"}:
+                    if node.func.attr[0].isupper():
+                        self.classes_used.add(node.func.attr)
+        elif isinstance(node.func, ast.Name):
+            if node.func.id in self.imports:
+                self.classes_used.add(node.func.id)
+
+
+if __name__ == "__main__":
+    for file in sys.argv[1:]:
+        with open(file) as fd:
+            content = fd.read()
+        tree = ast.parse(content)
+
+        import_visitor = ImportVisitor(file)
+        import_visitor.visit(tree)
+
+        visitor = CallVisitor(file, import_visitor.imports)
+        visitor.visit(tree)
+        for i in visitor.classes_used:
+            print(i)