FIX Fix line breaks in markdown tables

docs/changes.rst

@@ -16,6 +16,8 @@ v0.3
 - Add ``private`` as an optional argument to :meth:`.hub_utils.push` to
   optionally set the visibility status of a repo when pushing to the hub.
   :pr:`130` by `Adrin Jalali`_.
+- Fix a bug that resulted in markdown tables being rendered incorrectly if
+  entries contained line breaks by `Benjamin Bossan`_.
 
 v0.2
 ----

skops/card/_model_card.py

@@ -28,6 +28,19 @@ def wrap_as_details(text: str, folded: bool) -> str:
     return f"<details>\n<summary> Click to expand </summary>\n\n{text}\n\n</details>"
 
 
+def _clean_table(table: str) -> str:
+    # replace line breaks "\n" with html tag <br />, however, leave end-of-line
+    # line breaks (eol_lb) intact
+    eol_lb = "|\n"
+    placeholder = "$%!?"  # arbitrary sting that never appears naturally
+    table = (
+        table.replace(eol_lb, placeholder)
+        .replace("\n", "<br />")
+        .replace(placeholder, eol_lb)
+    )
+    return table
+
+
 @dataclass
 class PlotSection:
     """Adds a link to a figure to the model card"""
@@ -48,7 +61,7 @@ def __repr__(self) -> str:
 class TableSection:
     """Adds a table to the model card"""
 
-    table: dict["str", list[Any]]
+    table: dict[str, list[Any]]
     folded: bool = False
 
     def __post_init__(self) -> None:
@@ -78,8 +91,8 @@ def format(self) -> str:
         else:
             headers = self.table.keys()
 
-        table = tabulate(
-            self.table, tablefmt="github", headers=headers, showindex=False
+        table = _clean_table(
+            tabulate(self.table, tablefmt="github", headers=headers, showindex=False)
         )
         return wrap_as_details(table, folded=self.folded)
 
@@ -467,10 +480,12 @@ def _extract_estimator_config(self) -> str:
             Markdown table of hyperparameters.
         """
         hyperparameter_dict = self.model.get_params(deep=True)
-        return tabulate(
-            list(hyperparameter_dict.items()),
-            headers=["Hyperparameter", "Value"],
-            tablefmt="github",
+        return _clean_table(
+            tabulate(
+                list(hyperparameter_dict.items()),
+                headers=["Hyperparameter", "Value"],
+                tablefmt="github",
+            )
         )
 
     @staticmethod

skops/card/tests/test_card.py

@@ -68,6 +68,19 @@ def test_hyperparameter_table(destination_path, model_card):
     assert "fit_intercept" in model_card
 
 
+def test_hyperparameter_table_with_line_break(destination_path):
+    # Hyperparameters can contain values with line breaks, "\n", in them. In
+    # that case, the markdown table is broken. Check that the hyperparameter
+    # table we create properly replaces the "\n" with "<br />".
+    class EstimatorWithLbInParams:
+        def get_params(self, deep=False):
+            return {"fit_intercept": True, "n_jobs": "line\nwith\nbreak"}
+
+    model_card = Card(EstimatorWithLbInParams())
+    model_card = model_card.render()
+    assert "| n_jobs           | line<br />with<br />break         |" in model_card
+
+
 def test_plot_model(destination_path, model_card):
     model_card = model_card.render()
     assert "<style>" in model_card
@@ -415,3 +428,35 @@ def test_folded(self, table_dict, folded):
             assert "<details>" in output
         else:
             assert "<details>" not in output
+
+    def test_line_break_in_entry(self, table_dict):
+        # Line breaks are not allowed inside markdown tables, so check that
+        # they're removed. We test 3 conditions here:
+
+        # 1. custom object with line breaks in repr
+        # 2. string with line break in the middle
+        # 3. string with line break at start, middle, and end
+
+        # Note that for the latter, tabulate will automatically strip the line
+        # breaks from the start and end.
+        class LineBreakInRepr:
+            """Custom object whose repr has a line break"""
+
+            def __repr__(self) -> str:
+                return "obj\nwith lb"
+
+        table_dict["with break"] = [
+            LineBreakInRepr(),
+            "hi\nthere",
+            """
+entry with
+line breaks
+""",
+        ]
+        section = TableSection(table=table_dict)
+        expected = """|   split |   score | with break   |
+|---------|---------|--------------|
+|       1 |       4 | obj<br />with lb              |
+|       2 |       5 | hi<br />there              |
+|       3 |       6 | entry with<br />line breaks              |"""
+        assert section.format() == expected