lql-cli 0.6.0__tar.gz → 0.7.0__tar.gz

{lql_cli-0.6.0 → lql_cli-0.7.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: lql-cli
-Version: 0.6.0
+Version: 0.7.0
 Summary: lql — CLI for the Liquid DataViewer platform
 Project-URL: Homepage, https://github.com/Liquid4All/lql
 Author: Liquid AI
@@ -166,11 +166,23 @@ lql preview <file.jsonl|file.json>     Local file: each line/object is a row
 lql preview <dataset-id>               Platform dataset (fetched & paged lazily)
 lql preview <org/name> --hf            HuggingFace repo: sync to DataViewer, then view
 lql preview <src> -c <field>           Force field(s) as conversations (repeatable)
+lql preview <src> -f "col=value"       Filter rows (repeatable, AND); local & platform
 lql preview <src> -n <N>               Page size when paging a platform dataset
 lql preview <src> --offset N           Start at row index N
 lql preview <src> --title "<title>"    Title shown in the viewer header
 ```
 
+**Filtering (`--filter`/`-f`).** Show only matching rows — works on local files and
+platform datasets (platform filtering runs server-side). Repeatable; filters AND
+together; string match is case-insensitive. Operators: `=`, `!=`, `~` (contains),
+`>`, `<`, `>=`, `<=`.
+
+```
+lql preview <dataset-id> -f "domain=telecom"
+lql preview data.jsonl -f "reward>=0.8" -f "split=test"     # both must hold
+lql preview <dataset-id> -f "model~lfm"                      # contains
+```
+
 **HuggingFace datasets (`--hf`).** `lql preview org/name --hf` syncs the repo
 into a DataViewer workspace, then opens it. You pick the target workspace from
 an interactive list (or pass `--workspace <id>`; `--split` defaults to `train`).

{lql_cli-0.6.0 → lql_cli-0.7.0}/README.md

@@ -150,11 +150,23 @@ lql preview <file.jsonl|file.json>     Local file: each line/object is a row
 lql preview <dataset-id>               Platform dataset (fetched & paged lazily)
 lql preview <org/name> --hf            HuggingFace repo: sync to DataViewer, then view
 lql preview <src> -c <field>           Force field(s) as conversations (repeatable)
+lql preview <src> -f "col=value"       Filter rows (repeatable, AND); local & platform
 lql preview <src> -n <N>               Page size when paging a platform dataset
 lql preview <src> --offset N           Start at row index N
 lql preview <src> --title "<title>"    Title shown in the viewer header
 ```
 
+**Filtering (`--filter`/`-f`).** Show only matching rows — works on local files and
+platform datasets (platform filtering runs server-side). Repeatable; filters AND
+together; string match is case-insensitive. Operators: `=`, `!=`, `~` (contains),
+`>`, `<`, `>=`, `<=`.
+
+```
+lql preview <dataset-id> -f "domain=telecom"
+lql preview data.jsonl -f "reward>=0.8" -f "split=test"     # both must hold
+lql preview <dataset-id> -f "model~lfm"                      # contains
+```
+
 **HuggingFace datasets (`--hf`).** `lql preview org/name --hf` syncs the repo
 into a DataViewer workspace, then opens it. You pick the target workspace from
 an interactive list (or pass `--workspace <id>`; `--split` defaults to `train`).

{lql_cli-0.6.0 → lql_cli-0.7.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "lql-cli"
-version = "0.6.0"
+version = "0.7.0"
 description = "lql — CLI for the Liquid DataViewer platform"
 readme = "README.md"
 requires-python = ">=3.12"

lql_cli-0.7.0/src/lql/__init__.py

@@ -0,0 +1 @@
+__version__ = "0.7.0"

{lql_cli-0.6.0 → lql_cli-0.7.0}/src/lql/commands/instructions.py

@@ -100,8 +100,17 @@ DataViewer workspace (you pick one interactively, or pass --workspace <id>;
 --split defaults to train) and reused on later previews (dedup by repo+split).
 
 Options: -c/--column (field(s) to treat as conversations; default auto-detect,
-repeatable), -n/--limit (page size when paging a platform dataset), --offset
-(start row index), --title, --hf, --split, --workspace, --profile, --api-url.
+repeatable), -f/--filter (filter rows; see below), -n/--limit (page size when
+paging a platform dataset), --offset (start row index), --title, --hf, --split,
+--workspace, --profile, --api-url.
+
+Filtering: -f/--filter "col<op>value" shows only matching rows — works on local
+files and platform datasets (server-side for platform). Repeatable; filters AND
+together; string compare is case-insensitive. Operators: = (eq), != (ne),
+~ (contains), >, <, >=, <=.
+
+  lql preview <dataset-id> -f "domain=telecom" -f "reward>=0.8"
+  lql preview data.jsonl -f "model~lfm"
 
 Navigation: two modes toggled with m — pager (one sample at a time; ←/→ or
 n/b switch samples, ↑/↓/j/k scroll) and scroll (all samples; n/b jump between

{lql_cli-0.6.0 → lql_cli-0.7.0}/src/lql/commands/preview.py

@@ -759,6 +759,67 @@ def _choose_workspace(client: ApiClient, tui_mod) -> Optional[str]:
     return choice
 
 
+# --------------------------------------------------------------------------
+# Row filtering (--filter "col<op>value")
+# --------------------------------------------------------------------------
+
+# Maps each CLI symbol to the platform filter API's operator name (the same
+# names work server-side and locally). _parse_filters picks the earliest operator
+# (longest on a tie), so list order doesn't affect correctness.
+_FILTER_OPS = [(">=", "gte"), ("<=", "lte"), ("!=", "ne"), ("~", "contains"), ("=", "eq"), (">", "gt"), ("<", "lt")]
+_NUMERIC_OPS = {"gt": lambda c, v: c > v, "lt": lambda c, v: c < v, "gte": lambda c, v: c >= v, "lte": lambda c, v: c <= v}
+
+
+def _parse_filters(specs: Optional[List[str]]) -> List[tuple]:
+    """Parse ['col=value', 'reward>=0.5', 'name~kod'] → [(col, op, value), ...].
+
+    Splits on the EARLIEST operator (longest on a tie, so 'reward>=5' is gte not
+    gt), keeping operator chars in the value intact (e.g. 'q=a>b' → col 'q', value
+    'a>b'). Rejects an empty column or value."""
+    out: List[tuple] = []
+    for spec in specs or []:
+        chosen = None  # (index, symbol, op_name)
+        for sym, op in _FILTER_OPS:
+            i = spec.find(sym)
+            if i > 0 and (chosen is None or i < chosen[0] or (i == chosen[0] and len(sym) > len(chosen[1]))):
+                chosen = (i, sym, op)
+        if chosen is None:
+            print_error(
+                f"Invalid --filter '{spec}'. Use col=value, col!=value, col~text, or col>/</>=/<= N.",
+                "bad_filter",
+            )
+            raise typer.Exit(1)
+        i, sym, op = chosen
+        col, val = spec[:i].strip(), spec[i + len(sym):].strip()
+        if not col or not val:
+            print_error(f"Invalid --filter '{spec}': both a column and a value are required.", "bad_filter")
+            raise typer.Exit(1)
+        out.append((col, op, val))
+    return out
+
+
+def _cell_matches(cell: object, op: str, val: str) -> bool:
+    if op == "contains":
+        return cell is not None and val.lower() in str(cell).lower()
+    if op in ("eq", "ne"):
+        equal = cell is not None and str(cell).strip().lower() == val.strip().lower()
+        return equal if op == "eq" else not equal
+    try:
+        return _NUMERIC_OPS[op](float(cell), float(val))  # gt/lt/gte/lte
+    except (TypeError, ValueError):
+        return False
+
+
+def _row_matches(row: object, filters: List[tuple]) -> bool:
+    """Client-side predicate (local files). A non-dict row can't match a column
+    filter. All filters AND together."""
+    if not filters:
+        return True
+    if not isinstance(row, dict):
+        return False
+    return all(_cell_matches(row.get(col), op, val) for col, op, val in filters)
+
+
 # --------------------------------------------------------------------------
 # Command
 # --------------------------------------------------------------------------
@@ -772,6 +833,13 @@ def preview(
     ] = None,
     limit: Annotated[int, typer.Option("--limit", "-n", help="Page size when paging a platform dataset")] = 25,
     offset: Annotated[int, typer.Option("--offset", help="Start at this row index")] = 0,
+    filter_: Annotated[
+        Optional[List[str]],
+        typer.Option(
+            "--filter", "-f",
+            help="Filter rows: 'col=value', 'col!=value', 'col~text' (contains), or 'col>/</>=/<= N'. Repeatable (AND).",
+        ),
+    ] = None,
     title: Annotated[Optional[str], typer.Option("--title", help="Title shown in the viewer header")] = None,
     hf: Annotated[
         bool, typer.Option("--hf", help="Treat SOURCE as a HuggingFace repo (org/name); sync it to DataViewer, then view")
@@ -801,13 +869,20 @@ def preview(
         print_error("The terminal viewer requires 'textual'. Install it: pip install textual", "missing_textual")
         raise typer.Exit(1)
 
+    filters = _parse_filters(filter_)
     local_path = Path(source)
     is_local = (not hf) and local_path.exists() and local_path.is_file()
 
-    # Local file → load whole, view immediately.
+    # Local file → load whole (filter client-side), view immediately.
     if is_local:
+        rows = _load_local(local_path)
+        if filters:
+            rows = [r for r in rows if _row_matches(r, filters)]
+            if not rows:
+                print_error("No rows match the filter(s).", "no_match")
+                raise typer.Exit(3)
         tui_mod.run(
-            tui_mod.RowSource(initial=_load_local(local_path)),
+            tui_mod.RowSource(initial=rows),
             title or local_path.name,
             forced_cols=column or None,
             start_idx=max(0, offset),
@@ -834,11 +909,17 @@ def preview(
         view_title = title or f"dataset {source}"
 
     page_size = limit if limit and limit > 0 else 25
+    api_filters = [{"column": col, "operator": op, "value": val} for col, op, val in filters]
 
     def _fetch_page(off: int, lim: int) -> List[object]:
-        data = client.get(
-            f"/v1/datasets/{q(dataset_id)}/rows", params={"limit": str(lim), "offset": str(offset + off)}
-        ).json()
+        params = {"limit": str(lim), "offset": str(offset + off)}
+        if api_filters:
+            # Server-side filtering via the same endpoint `eval samples` uses.
+            data = client.post(
+                f"/v1/datasets/{q(dataset_id)}/rows/filter", json={"filters": api_filters}, params=params
+            ).json()
+        else:
+            data = client.get(f"/v1/datasets/{q(dataset_id)}/rows", params=params).json()
         return _normalize_loaded(data)
 
     # Fetch the first page up front (with feedback) so the viewer opens already
@@ -857,8 +938,13 @@ def preview(
             if first:
                 break
     if not first:
-        msg = f"Dataset returned no rows for split '{split}'." if hf else "Dataset returned no rows."
-        print_error(msg, "empty_dataset")
+        if filters:
+            msg, code = "No rows match the filter(s).", "no_match"
+        elif hf:
+            msg, code = f"Dataset returned no rows for split '{split}'.", "empty_dataset"
+        else:
+            msg, code = "Dataset returned no rows.", "empty_dataset"
+        print_error(msg, code)
         raise typer.Exit(3)
     row_source = tui_mod.RowSource(initial=first, fetch_page=_fetch_page, page_size=page_size)
     tui_mod.run(

{lql_cli-0.6.0 → lql_cli-0.7.0}/uv.lock

@@ -185,7 +185,7 @@ wheels = [
 
 [[package]]
 name = "lql-cli"
-version = "0.6.0"
+version = "0.7.0"
 source = { editable = "." }
 dependencies = [
     { name = "httpx" },

lql_cli-0.6.0/src/lql/__init__.py

@@ -1 +0,0 @@
-__version__ = "0.6.0"