aboutsummaryrefslogtreecommitdiff
path: root/src/exchangedb/exchange-0001.sql
diff options
context:
space:
mode:
authorChristian Grothoff <grothoff@gnunet.org>2022-03-27 10:32:28 +0200
committerChristian Grothoff <grothoff@gnunet.org>2022-03-27 10:32:28 +0200
commitd0a69da8954fd72f361795c2e007bad3fe5accd1 (patch)
tree821cc25e6f614ecfcd97d77bbc75ee24293887a7 /src/exchangedb/exchange-0001.sql
parent646c9ad0611e5320a460206dec4fdfd516dc8f64 (diff)
downloadexchange-d0a69da8954fd72f361795c2e007bad3fe5accd1.tar.xz
towards removing tiny bit
Diffstat (limited to 'src/exchangedb/exchange-0001.sql')
-rw-r--r--src/exchangedb/exchange-0001.sql69
1 files changed, 56 insertions, 13 deletions
diff --git a/src/exchangedb/exchange-0001.sql b/src/exchangedb/exchange-0001.sql
index b2fb52ac2..e6902ed1c 100644
--- a/src/exchangedb/exchange-0001.sql
+++ b/src/exchangedb/exchange-0001.sql
@@ -772,7 +772,7 @@ CREATE TABLE IF NOT EXISTS deposits_by_ready_default
CREATE TABLE IF NOT EXISTS deposits_for_matching
(refund_deadline INT8 NOT NULL
- ,shard INT8 NOT NULL
+ ,merchant_pub BYTEA NOT NULL CHECK (LENGTH(merchant_pub)=32)
,coin_pub BYTEA NOT NULL CHECK (LENGTH(coin_pub)=32) REFERENCES known_coins (coin_pub) ON DELETE CASCADE
,deposit_serial_id INT8
)
@@ -782,7 +782,7 @@ COMMENT ON TABLE deposits_for_matching
CREATE INDEX IF NOT EXISTS deposits_for_matching_main_index
ON deposits_for_matching
- (refund_deadline ASC, shard, coin_pub);
+ (refund_deadline ASC, merchant_pub, coin_pub);
CREATE TABLE IF NOT EXISTS deposits_for_matching_default
PARTITION OF deposits_for_matching
@@ -818,12 +818,12 @@ BEGIN
THEN
INSERT INTO deposits_for_matching
(refund_deadline
- ,shard
+ ,merchant_pub
,coin_pub
,deposit_serial_id)
VALUES
(NEW.refund_deadline
- ,NEW.shard
+ ,NEW.merchant_pub
,NEW.coin_pub
,NEW.deposit_serial_id);
END IF;
@@ -866,7 +866,7 @@ BEGIN
THEN
DELETE FROM deposits_for_matching
WHERE refund_deadline = OLD.refund_deadline
- AND shard = OLD.shard
+ AND merchant_pub = OLD.merchant_pub
AND coin_pub = OLD.coin_pub
AND deposit_serial_id = OLD.deposit_serial_id;
END IF;
@@ -887,12 +887,12 @@ BEGIN
THEN
INSERT INTO deposits_for_matching
(refund_deadline
- ,shard
+ ,merchant_pub
,coin_pub
,deposit_serial_id)
VALUES
(NEW.refund_deadline
- ,NEW.shard
+ ,NEW.merchant_pub
,NEW.coin_pub
,NEW.deposit_serial_id);
END IF;
@@ -930,7 +930,7 @@ BEGIN
THEN
DELETE FROM deposits_for_matching
WHERE refund_deadline = OLD.refund_deadline
- AND shard = OLD.shard
+ AND merchant_pub = OLD.merchant_pub
AND coin_pub = OLD.coin_pub
AND deposit_serial_id = OLD.deposit_serial_id;
END IF;
@@ -1040,21 +1040,64 @@ $$;
SELECT add_constraints_to_wire_out_partition('default');
+CREATE OR REPLACE FUNCTION wire_out_delete_trigger()
+ RETURNS trigger
+ LANGUAGE plpgsql
+ AS $$
+BEGIN
+ DELETE FROM aggregation_tracking
+ WHERE wtid_raw = OLD.wtid_raw;
+ RETURN OLD;
+END $$;
+COMMENT ON FUNCTION wire_out_delete_trigger()
+ IS 'Replicate reserve_out deletions into aggregation_tracking. This replaces an earlier use of an ON DELETE CASCADE that required a DEFERRABLE constraint and conflicted with nice partitioning.';
+
+CREATE TRIGGER wire_out_on_delete
+ AFTER DELETE
+ ON wire_out
+ FOR EACH ROW EXECUTE FUNCTION wire_out_delete_trigger();
+
+
+
+-- ------------------------------ aggregation_transient ----------------------------------------
+
+-- Note: this table is not yet used; it is designed
+-- to allow us to get rid of the 'tiny BOOL' and
+-- the associated need to look at tiny
+-- deposits repeatedly.
+CREATE TABLE IF NOT EXISTS aggregation_transient
+ (amount_val INT8 NOT NULL
+ ,amount_frac INT4 NOT NULL
+ ,wire_target_h_payto BYTEA CHECK (LENGTH(wire_target_h_payto)=32)
+ ,exchange_account_section TEXT NOT NULL
+ ,wtid_raw BYTEA NOT NULL CHECK (LENGTH(wtid_raw)=32)
+ )
+ PARTITION BY HASH (wire_target_h_payto);
+COMMENT ON TABLE aggregation_transient
+ IS 'aggregations currently happening (lacking wire_out, usually because the amount is too low); this table is not replicated';
+COMMENT ON COLUMN aggregation_transient.amount_val
+ IS 'Sum of all of the aggregated deposits (without deposit fees)';
+COMMENT ON COLUMN aggregation_transient.wtid_raw
+ IS 'identifier of the wire transfer';
+
+CREATE TABLE IF NOT EXISTS aggregation_transient_default
+ PARTITION OF aggregation_transient
+ FOR VALUES WITH (MODULUS 1, REMAINDER 0);
+
+
-- ------------------------------ aggregation_tracking ----------------------------------------
--- FIXME-URGENT: add colum coin_pub to select by coin_pub + deposit_serial_id for more efficient deposit lookup!?
--- Or which direction(s) is this table used? Is the partitioning sane??
CREATE TABLE IF NOT EXISTS aggregation_tracking
(aggregation_serial_id BIGINT GENERATED BY DEFAULT AS IDENTITY -- UNIQUE
,deposit_serial_id INT8 PRIMARY KEY -- REFERENCES deposits (deposit_serial_id) ON DELETE CASCADE
- ,wtid_raw BYTEA NOT NULL CONSTRAINT wire_out_ref REFERENCES wire_out(wtid_raw) ON DELETE CASCADE DEFERRABLE
+ ,wtid_raw BYTEA NOT NULL CHECK (LENGTH(wtid_raw)=32)
)
PARTITION BY HASH (deposit_serial_id);
COMMENT ON TABLE aggregation_tracking
IS 'mapping from wire transfer identifiers (WTID) to deposits (and back)';
COMMENT ON COLUMN aggregation_tracking.wtid_raw
- IS 'We first create entries in the aggregation_tracking table and then finally the wire_out entry once we know the total amount. Hence the constraint must be deferrable and we cannot use a wireout_uuid here, because we do not have it when these rows are created. Changing the logic to first INSERT a dummy row into wire_out and then UPDATEing that row in the same transaction would theoretically reduce per-deposit storage costs by 5 percent (24/~460 bytes).';
+ IS 'identifier of the wire transfer';
CREATE TABLE IF NOT EXISTS aggregation_tracking_default
PARTITION OF aggregation_tracking
@@ -1070,7 +1113,7 @@ BEGIN
EXECUTE FORMAT (
'ALTER TABLE aggregation_tracking_' || partition_suffix || ' '
'ADD CONSTRAINT aggregation_tracking_' || partition_suffix || '_aggregation_serial_id_key '
- 'UNIQUE (aggregation_serial_id) '
+ 'UNIQUE (aggregation_serial_id);'
);
END
$$;