scistackplotdb 0.1.58__tar.gz → 0.1.59__tar.gz

{scistackplotdb-0.1.58 → scistackplotdb-0.1.59}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: scistackplotdb
-Version: 0.1.58
+Version: 0.1.59
 Summary: scidb-backed plotting: load variables into long tables and plot them with scistackplot
 Author: SciStack Contributors
 License-Expression: MIT
@@ -17,10 +17,10 @@ Classifier: Programming Language :: Python :: 3.13
 Classifier: Topic :: Scientific/Engineering :: Visualization
 Classifier: Typing :: Typed
 Requires-Python: >=3.10
-Requires-Dist: sciduckdb==0.1.58
-Requires-Dist: scistack-db==0.1.58
-Requires-Dist: scistacklog==0.1.58
-Requires-Dist: scistackplot==0.1.58
+Requires-Dist: sciduckdb==0.1.59
+Requires-Dist: scistack-db==0.1.59
+Requires-Dist: scistacklog==0.1.59
+Requires-Dist: scistackplot==0.1.59
 Provides-Extra: dev
 Requires-Dist: matplotlib>=3.6; extra == 'dev'
 Requires-Dist: pytest-cov>=4.0; extra == 'dev'

{scistackplotdb-0.1.58 → scistackplotdb-0.1.59}/src/scistackplotdb/load.py

@@ -248,7 +248,7 @@ def is_container_type(data_type: str) -> bool:
 
     Why it is worth having at all: classifying by value means
     :func:`sample_column_value` pulls a whole EMG trace out of DuckDB per
-    muscle, only to discard it. Measured on the user's project (scidb.log
+    muscle, only to discard it. Measured on the user's project (scistack.log
     2026-09-15), that was 4.2 s for ``FilteredDelsys`` and 5.4 s for ``RawEMG``
     — ~8 s added to every panel open, by two variables that offer no groupings
     at all.
@@ -686,7 +686,7 @@ def _log_field_coverage(
     per field, when fewer than :data:`SPARSE_FIELD_SHARE` of its records hold
     one — naming the levels those records sit at, because a plot of such a
     field shows only them and reads as a broken grouping ("Dist_m showed one
-    group/one subject": 6 of 714 records held a number, scidb.log
+    group/one subject": 6 of 714 records held a number, scistack.log
     2026-10-15). The plot is right; the data is what to look at.
 
     Never raises: a diagnostic must not break a load.
@@ -727,6 +727,21 @@ def _log_field_coverage(
                 f", at {located}" if c.numeric and located else "",
                 layer=LAYER,
             )
+        # The same finding, once, in the scientist's journal: it is about
+        # their data, but the WARNs above repeat on every load (171 of them in
+        # one session's log, 2026-10-22). One note per variable and set of
+        # columns per session; a different set is a new finding.
+        sparse = [c for c in gaps if c.numeric < SPARSE_FIELD_SHARE * c.records]
+        if sparse:
+            names = sorted(str(c.column) for c in sparse)
+            shown = ", ".join(names[:6]) + (f", +{len(names) - 6} more" if len(names) > 6 else "")
+            Log.journal_note(
+                f"{variable}: {len(names)} column(s) hold numbers in only some of "
+                f"their records ({shown}); a plot of them shows only those records. "
+                f"If they hold text, that is expected; if they should be numbers, "
+                f"check what the loading function returned.",
+                key=f"field_coverage:{variable}:{'|'.join(names)}",
+            )
     except Exception as exc:  # noqa: BLE001 — diagnostic only
         Log.debug("field coverage of %s failed: %s", variable, exc, layer=LAYER)
 
@@ -1016,7 +1031,7 @@ def attach_variants(
                 "%d of %d record(s) are superseded but carry NO code or "
                 "run-option axis to tell them apart — plotting the current "
                 "ones only. This usually means their lineage was not recorded; "
-                "check scidb.log for 'NO variable input-binding source'",
+                "check scistack.log for 'NO variable input-binding source'",
                 n_superseded,
                 len(record_ids),
                 layer=LAYER,

{scistackplotdb-0.1.58 → scistackplotdb-0.1.59}/src/scistackplotdb/source.py

@@ -1019,7 +1019,7 @@ class ScidbSource(BaseSource):
             if ambiguous:
                 # Name what was discarded. When every ambiguous location holds
                 # the SAME set of labels, the grouping variable is almost
-                # certainly one whole table saved at every location (scidb.log
+                # certainly one whole table saved at every location (scistack.log
                 # run n1irqety, 2026-09-29: the demographics CSV read once per
                 # subject, so every subject "was" the first row, Digitimer).
                 values = right.groupby(on, dropna=False)[factor].agg(
@@ -1630,7 +1630,7 @@ class ScidbSource(BaseSource):
         expensive question, asked once per click by :meth:`groupable_columns`.
 
         Why the split. Answering both together cost ~8 s on every panel open of
-        the user's project (scidb.log 2026-09-15): `FilteredDelsys` 4.2 s and
+        the user's project (scistack.log 2026-09-15): `FilteredDelsys` 4.2 s and
         `RawEMG` 5.4 s, both offering nothing, because classifying a column
         meant fetching one of its EMG traces. Stage 4 made *that* cheap; this
         makes the per-column work happen only when someone asks for a specific
@@ -1879,7 +1879,7 @@ class ScidbSource(BaseSource):
         **Container-typed columns are refused from the schema alone**, before
         any value is read. Classifying them by value meant one whole EMG trace
         fetched per muscle and thrown away: 4.2 s for ``FilteredDelsys`` and
-        5.4 s for ``RawEMG`` on the user's project (scidb.log 2026-09-15), ~8 s
+        5.4 s for ``RawEMG`` on the user's project (scistack.log 2026-09-15), ~8 s
         added to every panel open by two variables that offer nothing. The type
         check only ever REFUSES — see :func:`~scistackplotdb.load.
         is_container_type` for why that asymmetry is what keeps it safe.