pf_app/setup_sql/01_schema.sql
Paul Trowbridge 496642c545 Stamp each log entry with what it did
The change log joined the whole forecast table on every open to total rows
it had just written -- 2.5M rows to report a few thousand, and the Baseline
page's row and value columns paid the same cost again.

The totals go onto pf.log at write time instead. This is not a cache that
can drift: a log entry's forecast rows never change once written, because
only the operation owning the logid inserts them and the only thing that
removes them is undo, which deletes the log row too. measure_cols records
which columns the figures are denominated in, since the value and units
roles can be reassigned in col_meta and the numbers would otherwise quietly
come to mean something else.

Stamping is best-effort and runs after the commit: a failure to record what
happened must not roll back the thing that happened. The loads return their
new log id to make it possible, the adjustments take theirs from the rows
they return, and apply_mode 'each' writes one entry per slice, so it is a
set rather than a single id.

?recount=1 does it the old way and writes back what it finds. Stored totals
cannot drift on their own, but nothing stops someone deleting forecast rows
by hand, and a stored figure has no way to notice -- so there is a way back,
which doubles as the backfill for entries written before the columns
existed.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-09-18 10:40:39 -04:00

175 lines
8.4 KiB
SQL

-- Pivot Forecast schema install
-- Run once against target database: psql -d <db> -f setup_sql/01_schema.sql
CREATE SCHEMA IF NOT EXISTS pf;
CREATE TABLE IF NOT EXISTS pf.source (
id serial PRIMARY KEY,
schema text NOT NULL,
tname text NOT NULL,
label text,
status text NOT NULL DEFAULT 'active', -- active | archived
default_layout jsonb, -- Perspective view config used as the per-source default
created_at timestamptz NOT NULL DEFAULT now(),
created_by text,
UNIQUE (schema, tname)
);
-- backfill columns for existing installs
ALTER TABLE pf.source ADD COLUMN IF NOT EXISTS default_layout jsonb;
-- pf.dim_period: run setup_sql/gen_dim_period.sql to create and populate
CREATE TABLE IF NOT EXISTS pf.col_meta (
id serial PRIMARY KEY,
source_id integer NOT NULL REFERENCES pf.source(id) ON DELETE CASCADE,
cname text NOT NULL,
label text,
role text NOT NULL DEFAULT 'ignore', -- dimension | value | units | date | ignore
is_key boolean NOT NULL DEFAULT false, -- true = usable in WHERE slice
dim_group text, -- groups functionally dependent columns
dim_period_col text, -- pf.dim_period column this dimension derives from
in_grain boolean NOT NULL DEFAULT false, -- true = column defines the display grain
opos integer,
UNIQUE (source_id, cname)
);
-- backfill columns for existing installs (must follow the CREATE above)
ALTER TABLE pf.col_meta ADD COLUMN IF NOT EXISTS dim_group text;
ALTER TABLE pf.col_meta ADD COLUMN IF NOT EXISTS dim_period_col text;
ALTER TABLE pf.col_meta ADD COLUMN IF NOT EXISTS in_grain boolean NOT NULL DEFAULT false;
CREATE TABLE IF NOT EXISTS pf.version (
id serial PRIMARY KEY,
source_id integer NOT NULL REFERENCES pf.source(id) ON DELETE RESTRICT,
name text NOT NULL,
description text,
status text NOT NULL DEFAULT 'open', -- open | closed
exclude_iters jsonb NOT NULL DEFAULT '["reference"]'::jsonb,
created_at timestamptz NOT NULL DEFAULT now(),
created_by text,
closed_at timestamptz,
closed_by text,
UNIQUE (source_id, name)
);
CREATE TABLE IF NOT EXISTS pf.log (
id bigserial PRIMARY KEY,
version_id integer NOT NULL REFERENCES pf.version(id) ON DELETE CASCADE,
pf_user text NOT NULL,
stamp timestamptz NOT NULL DEFAULT now(),
operation text NOT NULL, -- baseline | reference | scale | recode | clone
slice jsonb,
params jsonb,
note text,
tag text -- initiative label, e.g. 'reduce_spend'; groups
-- adjustments into a bridge from baseline to current
);
-- adding tags to an install that predates them
ALTER TABLE pf.log ADD COLUMN IF NOT EXISTS tag text;
CREATE INDEX IF NOT EXISTS log_tag_idx ON pf.log (tag) WHERE tag IS NOT NULL;
-- seed tags for loads that predate the column: a baseline/reference note is the
-- segment's name ('Open Orders', 'Prior Year'), which is exactly what tag holds.
-- Adjustment notes are free text, not labels, so they are left alone.
UPDATE pf.log
SET tag = note
WHERE TRUE
AND tag IS NULL
AND note IS NOT NULL
AND note <> ''
AND operation IN ('baseline', 'reference');
-- Display order for the pivot's segment and bucket columns.
--
-- The segment's display name in the pivot, falling back to tag then note.
--
-- Separate from both because those have jobs already -- tag groups adjustments
-- into initiatives for the bridge, note is free commentary -- and because the
-- label carries the sort order. Perspective orders column groups by the value
-- string, so a leading "01 - " is how ordering is expressed; putting that in the
-- note would put it in every note.
ALTER TABLE pf.log ADD COLUMN IF NOT EXISTS label text;
-- Vestigial, both of them. They held the ordinal when the "01 - " prefix was
-- computed for the pivot rather than typed into label and bucket: bucket_order
-- sequenced the bucket columns, log.seq the segments within them. Nothing reads
-- either now, and nothing writes them -- kept only because dropping a column is
-- not worth a migration to reclaim two that cost nothing.
ALTER TABLE pf.version ADD COLUMN IF NOT EXISTS bucket_order jsonb;
ALTER TABLE pf.log ADD COLUMN IF NOT EXISTS seq integer;
-- What a segment contributes to, independent of pf_iter.
--
-- pf_iter answers "can operations write to these rows"; bucket answers "does this
-- belong in the forecast number". Those are not the same question -- Open Orders is
-- loaded as reference so nothing adjusts it, yet it is part of the forecast -- so
-- neither can be derived from the other.
--
-- Free text with suggested values (Forecast / Prior Year / Prior Prior Year / Plan)
-- rather than an enum, so a new banner does not need a migration. Blank by default:
-- until a segment is labelled, the pivot falls back to showing its own name.
ALTER TABLE pf.log ADD COLUMN IF NOT EXISTS bucket text;
CREATE INDEX IF NOT EXISTS log_bucket_idx ON pf.log (bucket) WHERE bucket IS NOT NULL;
-- The names a row falls back to when nobody has named it, per scenario. Null
-- means "use the built-in", which is the DISPLAY DEFAULTS block in
-- lib/sql_generator.js. Read through a join at query time, not baked into
-- pf.sql: those templates are keyed on (source_id, operation) and shared by
-- every version of a source.
ALTER TABLE pf.version ADD COLUMN IF NOT EXISTS adjustment_segment text;
ALTER TABLE pf.version ADD COLUMN IF NOT EXISTS adjustment_bucket text;
ALTER TABLE pf.version ADD COLUMN IF NOT EXISTS unlabeled_load text;
-- What the entry did, stamped when it did it.
--
-- Not a cache: a log entry's forecast rows never change after it is written.
-- Only the operation that owns the logid inserts them, and the only thing that
-- removes them is undo, which deletes this row too -- so these totals are fixed
-- at write time rather than derived from something that can move underneath
-- them. The change log was joining the whole forecast table to recompute them
-- on every open, 2.5M rows to total a few thousand.
--
-- measure_cols records which columns they are denominated in, since the value
-- and units roles can be reassigned in col_meta and the numbers would otherwise
-- quietly come to mean something else.
ALTER TABLE pf.log ADD COLUMN IF NOT EXISTS row_count integer;
ALTER TABLE pf.log ADD COLUMN IF NOT EXISTS value_total double precision;
ALTER TABLE pf.log ADD COLUMN IF NOT EXISTS units_total double precision;
ALTER TABLE pf.log ADD COLUMN IF NOT EXISTS measure_cols jsonb;
-- Master data for a dim_group: one row per key value, with its sibling columns.
--
-- The source is transactional and often a view over all history, so deriving a
-- member list from it is both slow and wrong -- slow because it means scanning
-- millions of rows, wrong because it can only describe what was transacted and
-- has no way to say a part is discontinued or that a new one exists before it
-- has sold. This table is the app's own list, refreshed from the source but
-- curatable independently of it.
CREATE TABLE IF NOT EXISTS pf.dim_member (
source_id integer NOT NULL REFERENCES pf.source(id) ON DELETE CASCADE,
dim_group text NOT NULL, -- matches pf.col_meta.dim_group
key_value text NOT NULL, -- the is_key column's value
attrs jsonb NOT NULL DEFAULT '{}'::jsonb, -- the sibling columns
is_active boolean NOT NULL DEFAULT true,
source_seen boolean NOT NULL DEFAULT true, -- present in the source at last refresh
added_at timestamptz NOT NULL DEFAULT now(),
updated_at timestamptz NOT NULL DEFAULT now(),
refreshed_at timestamptz,
PRIMARY KEY (source_id, dim_group, key_value)
);
CREATE INDEX IF NOT EXISTS dim_member_active_idx
ON pf.dim_member (source_id, dim_group) WHERE is_active;
-- generated operation SQL per source, stored after col_meta is configured
CREATE TABLE IF NOT EXISTS pf.sql (
id serial PRIMARY KEY,
source_id integer NOT NULL REFERENCES pf.source(id) ON DELETE CASCADE,
operation text NOT NULL, -- get_data | baseline | reference | scale | recode | clone | undo
sql text NOT NULL,
generated_at timestamptz NOT NULL DEFAULT now(),
UNIQUE (source_id, operation)
);