counts are records, statistics see the zeros: n = count(obs_id) in every template, n_samples = tows sampled

The zero-filled rows of slice_bio.sql had been counted as observations (sardine larvae: 6,158 → 52,741 "in
view", the observations bars showing effort instead of catch, the pills reading 55,301). obs_id is NULL only
on a zero-filled row, so n = count(obs_id) restores every count to the records while mean / se / median keep
running over all sampled tows; picker.sql reports n_filled beside them and the bundle README says
"N observation rows (+ M zero-filled tows)".

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XJj3nb6xYaquerksg2vkYz

Author: bbest

sql/cruise.sql

@@ -1,4 +1,4 @@
-SELECT cruise_key, count(*) AS n, count(DISTINCT root_id) AS n_samples, count(DISTINCT grid_key) AS n_sta,
+SELECT cruise_key, count(obs_id) AS n, count(DISTINCT root_id) AS n_samples, count(DISTINCT grid_key) AS n_sta,
        avg({{val}}) AS mean, median({{val}}) AS med, epoch(min(datetime)) AS t0, epoch(max(datetime)) AS t1
 FROM slice
 WHERE cruise_key IS NOT NULL AND {{where}}

sql/cruise_samples.sql

@@ -1,6 +1,6 @@
 -- the chosen cruise's sampled values, one dot per root sample
 SELECT root_id, any_value(latitude) AS latitude, any_value(longitude) AS longitude, any_value(grid_key) AS grid_key,
-       count(*) AS n, avg({{val}}) AS mean, median({{val}}) AS med, epoch(min(datetime)) AS t
+       count(obs_id) AS n, avg({{val}}) AS mean, median({{val}}) AS med, epoch(min(datetime)) AS t
 FROM slice
 WHERE cruise_key = {{cruise}} AND latitude IS NOT NULL AND {{where}}
 GROUP BY root_id

sql/depth_strip.sql

@@ -1,5 +1,5 @@
 -- the water-column strip: median / IQR per 10 m bin over the current selection (all depths, so the brush has context)
-SELECT depth_bin, count(*) AS n, median({{val}}) AS med,
+SELECT depth_bin, count(obs_id) AS n, median({{val}}) AS med,
        quantile_cont({{val}}, 0.25) AS q1, quantile_cont({{val}}, 0.75) AS q3
 FROM slice
 WHERE depth_bin IS NOT NULL AND {{where_nodepth}}

sql/depth_strip_ds.sql

@@ -1,6 +1,6 @@
 -- the maximized water-column profile: the per-dataset median per 10 m bin (one dotted line per dataset,
 -- beside the all-dataset median + IQR of depth_strip.sql)
-SELECT dataset_key, depth_bin, count(*) AS n, median({{val}}) AS med,
+SELECT dataset_key, depth_bin, count(obs_id) AS n, median({{val}}) AS med,
        quantile_cont({{val}}, 0.25) AS q1, quantile_cont({{val}}, 0.75) AS q3
 FROM slice
 WHERE depth_bin IS NOT NULL AND {{where_nodepth}}

sql/hex.sql

@@ -1,6 +1,6 @@
 -- the hex column is printf('%x', parent(hex7, res)): an H3 parent is bit arithmetic on the res-7 cell
 -- (calcofi4db::h3_parent_sql), rendered as the standard H3 string deck.gl / h3-js take directly
-SELECT {{hex}} AS hex, count(*) AS n, count(DISTINCT root_id) AS n_samples,
+SELECT {{hex}} AS hex, count(obs_id) AS n, count(DISTINCT root_id) AS n_samples,
        avg({{val}}) AS mean, median({{val}}) AS med
 FROM slice
 WHERE hex7 IS NOT NULL AND {{where}}

sql/picker.sql

@@ -1,5 +1,7 @@
--- D8 rule 4: dataset x life stage x effort class x gear, with how many rows each denominator can serve
+-- D8 rule 4: dataset x life stage x effort class x gear, with how many RECORDS each denominator can serve
+-- (n_filled: the zero-filled tows of slice_bio.sql beside them, never counted as observations)
 SELECT dataset_key, life_stage, effort_class, tow_type, any_value(units) AS units,
-       count(*) AS n, count(density_per_10m2) AS n_10m2, count(density_per_1000m3) AS n_1000m3,
+       count(obs_id) AS n, count(density_per_10m2) FILTER (WHERE obs_id IS NOT NULL) AS n_10m2, count(density_per_1000m3) FILTER (WHERE obs_id IS NOT NULL) AS n_1000m3,
+       count(*) FILTER (WHERE obs_id IS NULL) AS n_filled,
        count(*) FILTER (WHERE NOT qual_ok) AS n_flagged
 FROM slice GROUP BY ALL ORDER BY 1, 2, 3, 4

sql/region.sql

@@ -1,5 +1,5 @@
 -- exact per-root-sample membership (sample_spatial), one layer at a time because layers overlap
-SELECT sp.spatial_key, any_value(sp.spatial_name) AS spatial_name, count(*) AS n, count(DISTINCT s.root_id) AS n_samples,
+SELECT sp.spatial_key, any_value(sp.spatial_name) AS spatial_name, count(s.obs_id) AS n, count(DISTINCT s.root_id) AS n_samples,
        avg({{val}}) AS mean, median({{val}}) AS med, min(year) AS y0, max(year) AS y1
 FROM slice s JOIN {{spatial_src}} sp USING (root_id)
 WHERE sp.layer = {{layer}} AND {{where}}

sql/section_bio.sql

@@ -1,6 +1,6 @@
 -- bio has no depth axis in this cut (tows are depth-integrated): the section is line x year
 SELECT station, year,
-       avg({{val}}) AS v, count(*) AS n
+       avg({{val}}) AS v, count(obs_id) AS n
 FROM slice
 WHERE line = {{line}} AND {{where}}
 GROUP BY ALL ORDER BY station, year

sql/section_cruises.sql

@@ -1,6 +1,6 @@
 -- cruises that sampled this line, newest first (a YYYY-MM-NODC key sorts chronologically by itself);
 -- n_sta is the picker's bar and App.tsx picks the cruise with the most stations as the default
-SELECT cruise_key, count(DISTINCT grid_key) AS n_sta, count(*) AS n, min(year) AS year
+SELECT cruise_key, count(DISTINCT grid_key) AS n_sta, count(obs_id) AS n, min(year) AS year
 FROM slice
 WHERE cruise_key IS NOT NULL AND line = {{line}} AND {{where}}
 GROUP BY cruise_key ORDER BY cruise_key DESC

sql/slice_bio.sql

@@ -5,7 +5,8 @@
 -- dataset sampled (a root sample with >= 1 row of any taxon; the effort it can standardize by is what its
 -- rows carry) gets a zero row per life stage the dataset records this taxon at, unless it has a row already.
 -- a dataset that ships its own zeros (cufes, zooscan, zoodb, phyllosoma) is left as it is. zero rows have
--- obs_id NULL and qual_ok TRUE.
+-- obs_id NULL and qual_ok TRUE; every template counts n = count(obs_id) (records) and n_samples = tows, so the
+-- counts shown stay the records and only the statistics (mean, se, median) see the zeros.
 CREATE OR REPLACE TABLE slice AS
 WITH pos AS (
   SELECT obs_id, dataset_key, root_id, grid_key,