tui: fix ColumnsModal crash on slugs with duplicate column names (0.1.5)

The new Columns modal in 0.1.4 used the bare column name as the Textual
DataTable row key, which crashes with DuplicateKey on the 11 slugs that
ship parquet schemas with legitimately repeated top-level names — the
osmi-mental-health-in-tech-* survey series (2016–2023, repeating "Why
or why not?" follow-ups under each yes/no item), uci-spambase, uci-
parkinsons, and uk-price-paid. The by-name stats dict silently
collapsed duplicate-named entries on top of that.

Suffix repeated names with " (2)", " (3)", etc. at construction time so
the DataTable row key is unique and stats_by_name keeps each occurrence
separately. The underlying parquet column names are untouched; the
suffix is display + lookup only.

The deeper read_column_stats over-counting bug (length summed across
duplicate-named fields, min/max/nulls dropped because the len(indices)
== 1 branch is skipped) is left as a separate follow-up — fixing it
would change snapshot.json contents for the affected slugs.

Co-Authored-By: Claude <noreply@anthropic.com>
Signed-off-by: mprammer <martin@spiraldb.com>

Author: mprammer

CHANGELOG.md

@@ -5,6 +5,23 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.1.5] - 2026-05-17
+
+### Fixed
+
+- **TUI Columns modal crash on slugs with duplicate column names.**
+  Eleven slugs ship parquet schemas with legitimately repeated
+  top-level column names — the `osmi-mental-health-in-tech-*` survey
+  series (2016 through 2023) repeats "Why or why not?" follow-ups
+  under each yes/no item, and `uci-spambase`, `uci-parkinsons`, and
+  `uk-price-paid` each have one or more repeated headers. The new
+  Columns modal used the bare column name as the Textual DataTable
+  row key, so the second occurrence crashed with `DuplicateKey`.
+  Repeated names are now suffixed with ` (2)`, ` (3)`, etc. for
+  display + lookup; the by-name stats dict no longer silently
+  collapses entries either. The underlying parquet's column names
+  are unchanged.
+
 ## [0.1.4] - 2026-05-17
 
 ### Added

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "raincloud"
-version = "0.1.4"
+version = "0.1.5"
 description = "Client-reproducible pipeline for building a curated catalog of public datasets as Parquet + Vortex files."
 readme = "README.md"
 requires-python = ">=3.11"

scripts/pipeline/browse.py

@@ -1022,6 +1022,23 @@ class _DatasetModal(ModalScreen):
     ]
 
 
+def _dedupe_stat_names(stats: list[dict]) -> list[dict]:
+    """Suffix repeated `name` values with ` (2)`, ` (3)`, etc. so each entry is uniquely keyable."""
+    seen: dict[str, int] = {}
+    out: list[dict] = []
+    for s in stats:
+        name = s.get("name")
+        count = seen.get(name, 0) + 1
+        seen[name] = count
+        if count == 1:
+            out.append(s)
+        else:
+            new = dict(s)
+            new["name"] = f"{name} ({count})"
+            out.append(new)
+    return out
+
+
 class ColumnsModal(_DatasetModal):
     """Full per-column metadata for one slug. When the local parquet isn't
     built, falls back to docs/v1/snapshot.json so the modal can still show
@@ -1036,13 +1053,16 @@ def __init__(self, slug: str, spec: dict,
         super().__init__()
         self.slug = slug
         self.spec = spec
-        self.stats = stats
+        # Some slugs (osmi-* surveys, uci-spambase, uk-price-paid) carry
+        # legitimately duplicated top-level column names. Suffix repeats so
+        # the DataTable row key is unique and stats_by_name doesn't collapse.
+        self.stats = _dedupe_stat_names(stats) if stats is not None else None
         self.source = source  # "parquet" | "snapshot" | None
         # profile.json keyed-by-column-name. Empty when the profile stage
         # hasn't been run; the right detail pane then renders "no profile".
         self.profile_columns = (profile or {}).get("columns") or {}
         # parquet schema stats keyed by name for O(1) lookup from the detail pane.
-        self.stats_by_name = {s["name"]: s for s in (stats or [])}
+        self.stats_by_name = {s["name"]: s for s in (self.stats or [])}
 
     def compose(self) -> ComposeResult:
         suffix = " [dim](from snapshot)[/dim]" if self.source == "snapshot" else ""

tests/test_browse.py

@@ -335,6 +335,37 @@ def test_columns_modal_renders_built_state():
     assert m.profile_columns == {}
 
 
+def test_columns_modal_dedupes_duplicate_column_names():
+    """Survey slugs (osmi-* series) and a few others ship parquet schemas
+    with repeated top-level column names. The modal must suffix repeats
+    so the DataTable row key is unique and stats_by_name doesn't silently
+    collapse duplicate-named entries onto the last one."""
+    pytest.importorskip("textual")
+    from scripts.pipeline.browse import ColumnsModal
+
+    stats = [
+        {"name": "Q1", "type": "string", "length": 100,
+         "null_count": 0, "min": None, "max": None},
+        {"name": "Why or why not?", "type": "string", "length": 80,
+         "null_count": 0, "min": None, "max": None},
+        {"name": "Q2", "type": "string", "length": 90,
+         "null_count": 0, "min": None, "max": None},
+        {"name": "Why or why not?", "type": "string", "length": 75,
+         "null_count": 0, "min": None, "max": None},
+        {"name": "Why or why not?", "type": "string", "length": 60,
+         "null_count": 0, "min": None, "max": None},
+    ]
+    m = ColumnsModal("x", {"slug": "x"}, stats)
+    assert [s["name"] for s in m.stats] == [
+        "Q1", "Why or why not?", "Q2",
+        "Why or why not? (2)", "Why or why not? (3)",
+    ]
+    # stats_by_name keeps a distinct entry for each occurrence.
+    assert m.stats_by_name["Why or why not?"]["length"] == 80
+    assert m.stats_by_name["Why or why not? (2)"]["length"] == 75
+    assert m.stats_by_name["Why or why not? (3)"]["length"] == 60
+
+
 def test_render_column_detail_dtype_shapes():
     """`_render_column_detail` produces shape-appropriate multi-line markup."""
     pytest.importorskip("textual")