Add column metadata panel with stats

src/dt_browser/browser.py

@@ -28,6 +28,7 @@
     SelectFromTable,
 )
 from dt_browser.bookmarks import Bookmarks
+from dt_browser.column_metadata import ColumnMetadata
 from dt_browser.column_selector import ColumnSelector
 from dt_browser.custom_table import CustomTable, _color_name, polars_list_to_string
 from dt_browser.expression_box import ExpressionBox
@@ -248,9 +249,8 @@ def compute_active_search_idx_display(self):
 class RowDetail(Widget, can_focus=False, can_focus_children=False):
     DEFAULT_CSS = """
 RowDetail {
-    width: auto;
-    max_width: 50%;
-    min_width: 30%;
+    width: 100%;
+    height: 1fr;
     padding: 0 1;
     border: tall $primary;
 }
@@ -288,14 +288,43 @@ def watch_row_df(self):
         assert self._schema is not None
         display_df = display_df.join(self._schema, on=["Field"]).select(["Field", "dtype", "Value"])
         self._dt.set_dt(display_df, display_df.with_row_index(name=INDEX_COL).select([INDEX_COL]))
-        self.styles.width = self._dt.virtual_size.width + self.gutter.width + 1
         self._dt.refresh()
+        if isinstance(self.parent, DetailPanel):
+            self.parent.update_width()
         # self._dt.go_to_cell(coord)
 
+    @property
+    def content_width(self) -> int:
+        return self._dt.virtual_size.width + self.gutter.width + 1
+
     def compose(self):
         yield self._dt
 
 
+class DetailPanel(Widget, can_focus=False):
+    DEFAULT_CSS = """
+DetailPanel {
+    max-width: 50%;
+    min-width: 30%;
+    layout: vertical;
+}
+"""
+
+    def __init__(self, row_detail: RowDetail, column_metadata: ColumnMetadata, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self._row_detail = row_detail
+        self._column_metadata = column_metadata
+
+    def update_width(self) -> None:
+        row_detail_width = self._row_detail.content_width if not self._row_detail.row_df.is_empty() else 0
+        meta_width = self._column_metadata.content_size.width + self._column_metadata.gutter.width
+        self.styles.width = max(row_detail_width, meta_width)
+
+    def compose(self):
+        yield self._row_detail
+        yield self._column_metadata
+
+
 def from_file_path(path: pathlib.Path, has_header: bool = True) -> pl.DataFrame:
 
     if path.suffix in [".arrow", ".feather"]:
@@ -342,6 +371,7 @@ class DtBrowser(Widget):  # pylint: disable=too-many-public-methods,too-many-ins
     current_filter = reactive[str | None](None)
 
     cur_row = reactive(0)
+    cur_col = reactive(0)
     cur_total_rows = reactive(0)
     total_rows = reactive(0)
 
@@ -400,6 +430,9 @@ def __init__(
         self._ts_col_selector.styles.width = 1
 
         self._row_detail = RowDetail()
+        self._column_metadata = ColumnMetadata()
+        self._column_metadata.set_source_df(self._filtered_dt)
+        self._detail_panel = DetailPanel(self._row_detail, self._column_metadata)
 
         self._color_by_cache: LRUCache[tuple[str, ...], pl.Series] = LRUCache(5)
         self._last_message_ts = time.time()
@@ -623,10 +656,10 @@ async def action_show_save(self):
 
     async def watch_show_row_detail(self):
         if not self.show_row_detail:
-            if existing := self.query(RowDetail):
+            if existing := self.query(DetailPanel):
                 existing.remove()
         elif not self._display_dt.is_empty():
-            await self.query_one("#main_hori", Horizontal).mount(self._row_detail)
+            await self.query_one("#main_hori", Horizontal).mount(self._detail_panel)
 
     async def action_show_bookmarks(self):
         await self.mount(self._bookmarks, before=self.query_one(TableFooter))
@@ -644,6 +677,8 @@ async def action_timestamp_selector(self):
     def _set_filtered_dt(self, filtered_dt: pl.DataFrame, filtered_meta: pl.DataFrame, **kwargs):
         self._filtered_dt = filtered_dt
         self._meta_dt = filtered_meta
+        self._column_metadata.set_source_df(self._filtered_dt)
+        self._column_metadata.invalidate_cache()
         self._set_active_dt(self._filtered_dt, **kwargs)
 
     def _set_active_dt(self, active_dt: pl.DataFrame, new_row: int | None = None):
@@ -714,10 +749,20 @@ def enable_select_from_table(self, event: SelectFromTable):
     @on(CustomTable.CellHighlighted, selector="#main_table")
     async def handle_cell_highlight(self, event: CustomTable.CellHighlighted):
         self.cur_row = event.coordinate.row
+        col = event.coordinate.column
+        if col != self.cur_col:
+            self.cur_col = col
 
     def watch_cur_row(self):
         self._row_detail.row_df = self._display_dt[self.cur_row]
 
+    def watch_cur_col(self):
+        if self._display_dt.is_empty() or self.cur_col >= len(self._display_dt.columns):
+            return
+        col_name = self._display_dt.columns[self.cur_col]
+        dtype = self._display_dt.schema[col_name]
+        self._column_metadata.column_info = (col_name, dtype)
+
     @on(CustomTable.CellSelected, selector="#main_table")
     def handle_cell_select(self, event: CustomTable.CellSelected):
         if self._select_interest:
@@ -857,6 +902,9 @@ def on_mount(self):
         self.cur_total_rows = len(self._display_dt)
         self.total_rows = len(self._original_dt)
         self._row_detail.row_df = self._display_dt[0]
+        if not self._display_dt.is_empty():
+            col_name = self._display_dt.columns[0]
+            self._column_metadata.column_info = (col_name, self._display_dt.schema[col_name])
         if self.removed_cols:
             err_str = ", ".join(f"{k}: {v}" for k, v in self.removed_cols.items())
             self.notify(

src/dt_browser/column_metadata.py

@@ -0,0 +1,134 @@
+import polars as pl
+from rich.table import Table as RichTable
+from textual import work
+from textual.reactive import reactive
+from textual.widget import Widget
+from textual.widgets import Static
+
+
+def _categorical_stats(series: pl.Series) -> list[tuple[str, str]]:
+    n_unique = series.n_unique()
+    stats: list[tuple[str, str]] = [("Unique values", str(n_unique))]
+    val_col = series.name
+    vc = series.value_counts().sort(["count", val_col], descending=[True, False]).head(10)
+    for row in vc.iter_rows(named=True):
+        stats.append((f"  {row[val_col]}", str(row["count"])))
+    return stats
+
+
+def _numeric_stats(series: pl.Series) -> list[tuple[str, str]]:
+    s = series.drop_nulls()
+    if s.is_empty():
+        return [("", "No data")]
+    stats = [
+        ("Min", str(s.min())),
+        ("Q1", str(s.quantile(0.25))),
+        ("Median", str(s.median())),
+        ("Q3", str(s.quantile(0.75))),
+        ("Max", str(s.max())),
+    ]
+    if s.dtype.is_float():
+        nan_count = s.is_nan().sum()
+        if nan_count > 0:
+            stats.append(("NaN", str(nan_count)))
+    return stats
+
+
+def _temporal_stats(series: pl.Series) -> list[tuple[str, str]]:
+    s = series.drop_nulls()
+    if s.is_empty():
+        return [("", "No data")]
+    return [
+        ("Min", str(s.min())),
+        ("Max", str(s.max())),
+    ]
+
+
+def _boolean_stats(series: pl.Series) -> list[tuple[str, str]]:
+    true_count = series.sum()
+    null_count = series.null_count()
+    false_count = len(series) - (true_count or 0) - null_count
+    return [
+        ("True", str(true_count)),
+        ("False", str(false_count)),
+    ]
+
+
+def compute_column_stats(series: pl.Series) -> list[tuple[str, str]]:
+    dtype = series.dtype
+    if dtype == pl.Categorical:
+        stats = _categorical_stats(series)
+    elif dtype.is_numeric():
+        stats = _numeric_stats(series)
+    elif dtype.is_temporal():
+        stats = _temporal_stats(series)
+    elif dtype.is_(pl.Boolean):
+        stats = _boolean_stats(series)
+    else:
+        return []
+    null_count = series.null_count()
+    if null_count > 0:
+        stats.append(("Null", str(null_count)))
+    return stats
+
+
+class ColumnMetadata(Widget, can_focus=False, can_focus_children=False):
+    DEFAULT_CSS = """
+ColumnMetadata {
+    width: 100%;
+    height: auto;
+    padding: 0 1;
+    border: tall $primary;
+}
+"""
+    column_info: reactive[tuple[str, pl.DataType] | None] = reactive(None)
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        self.border_title = "Column Metadata"
+        self._source_df: pl.DataFrame = pl.DataFrame()
+        self._cache: dict[str, list[tuple[str, str]]] = {}
+        self._static = Static("")
+
+    def set_source_df(self, df: pl.DataFrame) -> None:
+        self._source_df = df
+
+    def invalidate_cache(self) -> None:
+        self._cache.clear()
+
+    def _render_stats(self, col_name: str, stats: list[tuple[str, str]]) -> None:
+        self.border_title = f"Column: {col_name}"
+        if not stats:
+            self._static.update("")
+            return
+        table = RichTable(show_header=False, box=None, padding=(0, 1), expand=True)
+        table.add_column("Stat", no_wrap=True)
+        table.add_column("Value", no_wrap=True, justify="right")
+        for label, value in stats:
+            table.add_row(label, value)
+        self._static.update(table)
+        if self.parent is not None and hasattr(self.parent, "update_width"):
+            self.parent.update_width()
+
+    def watch_column_info(self) -> None:
+        if self.column_info is None or self._source_df.is_empty():
+            return
+        col_name, _ = self.column_info
+        if col_name not in self._source_df.columns:
+            return
+        if col_name in self._cache:
+            self._render_stats(col_name, self._cache[col_name])
+        else:
+            self.border_title = f"Column: {col_name}"
+            self._static.update("Computing...")
+            self._compute_stats(col_name)
+
+    @work(exclusive=True)
+    async def _compute_stats(self, col_name: str) -> None:
+        series = self._source_df[col_name]
+        stats = compute_column_stats(series)
+        self._cache[col_name] = stats
+        self._render_stats(col_name, stats)
+
+    def compose(self):
+        yield self._static