Misc: Weight enum, testing improvements

dp_wizard/shiny/components/icons.py

@@ -1,5 +1,7 @@
 from faicons import icon_svg
 
+# Find more icons on Font Awesome: https://fontawesome.com/search?ic=free
+
 data_source_icon = icon_svg("file")
 unit_of_privacy_icon = icon_svg("shield-halved")
 product_icon = icon_svg("cart-shopping")

dp_wizard/shiny/panels/analysis_panel/column_module.py

@@ -16,7 +16,7 @@
     only_for_screenreader,
     tutorial_box,
 )
-from dp_wizard.types import AnalysisName, ColumnName, Product
+from dp_wizard.types import AnalysisName, ColumnName, Product, Weight
 from dp_wizard.utils.code_generators import make_column_config_block
 from dp_wizard.utils.code_generators.analyses import (
     get_analysis_by_name,
@@ -29,7 +29,6 @@
 from dp_wizard.utils.shared import plot_bars
 
 default_analysis_type = histogram.name
-default_weight = "2"
 label_width = "10em"  # Just wide enough so the text isn't trucated.
 
 
@@ -141,7 +140,7 @@ def column_server(
     lower_bounds: reactive.Value[dict[ColumnName, float]],
     upper_bounds: reactive.Value[dict[ColumnName, float]],
     bin_counts: reactive.Value[dict[ColumnName, int]],
-    weights: reactive.Value[dict[ColumnName, str]],
+    weights: reactive.Value[dict[ColumnName, Weight]],
     is_tutorial_mode: reactive.Value[bool],
     is_sample_csv: bool,
     is_single_column: bool,
@@ -150,7 +149,10 @@ def column_server(
     def _set_hidden_inputs():
         # TODO: Is isolate still needed?
         with reactive.isolate():  # Without isolate, there is an infinite loop.
-            ui.update_numeric("weight", value=int(weights().get(name, default_weight)))
+            ui.update_numeric(
+                "weight",
+                value=int(weights().get(name, Weight.DEFAULT).value),
+            )
 
     @reactive.effect
     @reactive.event(input.analysis_type)
@@ -187,15 +189,15 @@ def _set_bins():
     @reactive.effect
     @reactive.event(input.weight)
     def _set_weight():
-        weights.set({**weights(), name: input.weight()})
+        weights.set({**weights(), name: Weight(input.weight())})
 
     @reactive.calc()
     def accuracy_histogram():
         lower_x = float(input.lower_bound())
         upper_x = float(input.upper_bound())
         bin_count = int(input.bins())
         weight = float(input.weight())
-        weights_sum = sum(float(weight) for weight in weights().values())
+        weights_sum = sum(float(weight.value) for weight in weights().values())
         info(f"Weight ratio for {name}: {weight}/{weights_sum}")
         if weights_sum == 0:
             # This function is triggered when column is removed;
@@ -349,12 +351,8 @@ def optional_weight_ui():
                 ui.input_select(
                     "weight",
                     "Weight",
-                    choices={
-                        "1": "Less accurate",
-                        default_weight: "Default",
-                        "4": "More accurate",
-                    },
-                    selected=default_weight,
+                    choices={w.value: str(w) for w in Weight},
+                    selected=Weight.DEFAULT.value,
                     width=label_width,
                 ),
                 tutorial_box(

dp_wizard/shiny/panels/results_panel/__init__.py

@@ -264,7 +264,7 @@ def analysis_plan() -> AnalysisPlan:
                     lower_bound=lower_bounds()[col],
                     upper_bound=upper_bounds()[col],
                     bin_count=int(bin_counts()[col]),
-                    weight=int(weights()[col]),
+                    weight=int(weights()[col].value),
                 )
             ]
             for col in weights().keys()

dp_wizard/types.py

@@ -5,6 +5,22 @@
 from shiny import reactive
 
 
+class Weight(Enum):
+    """
+    >>> print(Weight.MORE_ACCURATE)
+    More accurate
+    >>> ints = [int(w.value) for w in Weight]
+    >>> assert ints[2] / ints[1] == ints[1] / ints[0]
+    """
+
+    MORE_PRIVATE = "1"
+    DEFAULT = "2"
+    MORE_ACCURATE = "4"
+
+    def __str__(self) -> str:
+        return self.name.replace("_", " ").capitalize()
+
+
 class Product(Enum):
     STATISTICS = auto()
     SYNTHETIC_DATA = auto()
@@ -109,7 +125,7 @@ class AppState:
     lower_bounds: reactive.Value[dict[ColumnName, float]]
     upper_bounds: reactive.Value[dict[ColumnName, float]]
     bin_counts: reactive.Value[dict[ColumnName, int]]
-    weights: reactive.Value[dict[ColumnName, str]]
+    weights: reactive.Value[dict[ColumnName, Weight]]
     analysis_errors: reactive.Value[dict[ColumnName, bool]]
 
     # Release state:

tests/utils/test_code_generators.py

@@ -107,7 +107,7 @@ def test_make_column_config_block_for_histogram():
     )
 
 
-abc_csv = "tests/fixtures/abc.csv"
+abc_csv_path = str((package_root.parent / "tests/fixtures/abc.csv").absolute())
 
 
 def number_lines(text: str):
@@ -141,22 +141,17 @@ def number_lines(text: str):
 
 
 def id_for_plan(plan: AnalysisPlan):
-    columns = ", ".join(f"{v[0].analysis_name} of {k}" for k, v in plan.columns.items())
-    description = (
-        f"{plan.product} for {columns}; "
-        f"grouped by ({', '.join(plan.groups) or 'nothing'})"
-    )
-    return re.sub(r"\W+", "_", description)  # For selection with "pytest -k substring"
+    return re.sub(r"\W+", "_", str(plan))  # For selection with "pytest -k substring"
 
 
-plans = [
+plans_all_combos = [
     AnalysisPlan(
         product=product,
         groups=groups,
         columns=columns,
         contributions=contributions,
         contributions_entity="Family",
-        csv_path=abc_csv,
+        csv_path=abc_csv_path,
         epsilon=1,
         max_rows=100_000,
     )
@@ -178,6 +173,13 @@ def id_for_plan(plan: AnalysisPlan):
 ]
 
 
+# The matrix is very redundant! A subsample is sufficient,
+# but make sure it's relatively prime so we have coverage.
+mod = 7
+assert len(plans_all_combos) % mod != 0
+plans = [plan for i, plan in enumerate(plans_all_combos) if i % 7 == 0]
+
+
 expected_urls = [
     "https://docs.opendp.org/",
     "https://github.com/opendp/dp-wizard",
@@ -230,6 +232,7 @@ def test_make_notebook(plan):
 @pytest.mark.parametrize("plan", plans, ids=id_for_plan)
 def test_make_script(plan):
     script = ScriptGenerator(plan, "Note goes here!").make_py()
+    print(number_lines(script))
 
     # Make sure jupytext formatting doesn't bleed into the script.
     # https://jupytext.readthedocs.io/en/latest/formats-scripts.html#the-light-format
@@ -241,6 +244,6 @@ def test_make_script(plan):
         fp.flush()
 
         result = subprocess.run(
-            ["python", fp.name, "--csv", abc_csv], capture_output=True
+            ["python", fp.name, "--csv", abc_csv_path], capture_output=True
         )
         assert result.returncode == 0