Rework <apply-entities to be 40% faster and not blow the stack in CLJS.
* Batch up datoms into a smaller number of queries, improving transact speed by about 50%. * Restore transacting FTS attributes. * Implement retraction of freetext datoms.
This commit is contained in:
Richard Newman
parent
9136ba7425
commit
86b5a8ea8a
3 changed files with 357 additions and 162 deletions
|
|
@ -37,6 +37,9 @@
|
|||
(uncaughtException [_ thread ex]
|
||||
(println ex "Uncaught exception on" (.getName thread))))))
|
||||
|
||||
(def max-sql-vars 999) ;; TODO: generalize.
|
||||
|
||||
|
||||
;; ----------------------------------------------------------------------------
|
||||
;; define data-readers to be made available to EDN readers. in CLJS
|
||||
;; they're magically available. in CLJ, data_readers.clj may or may
|
||||
|
|
@ -96,7 +99,7 @@
|
|||
|
||||
(<av
|
||||
[db a v]
|
||||
"Search for datoms using the AVET index.")
|
||||
"Search for a single matching datom using the AVET index.")
|
||||
|
||||
(<apply-entities
|
||||
[db tx entities]
|
||||
|
|
@ -164,6 +167,324 @@
|
|||
:table-alias source/gensym-table-alias
|
||||
:make-constraints nil}))
|
||||
|
||||
(defn- retractions->queries [retractions tx fulltext? ->SQLite]
|
||||
(let
|
||||
[f-q
|
||||
"WITH vv AS (SELECT rowid FROM fulltext_values WHERE text = ?)
|
||||
INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0, sv, svalue_type_tag)
|
||||
VALUES (?, ?, (SELECT rowid FROM vv), ?, 0, ?, (SELECT rowid FROM vv), ?)"
|
||||
|
||||
non-f-q
|
||||
"INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0, sv, svalue_type_tag)
|
||||
VALUES (?, ?, ?, ?, 0, ?, ?, ?)"]
|
||||
(map
|
||||
(fn [[_ e a v]]
|
||||
(let [[v tag] (->SQLite a v)]
|
||||
(if (fulltext? a)
|
||||
[f-q
|
||||
v e a tx tag tag]
|
||||
[non-f-q
|
||||
e a v tx tag v tag])))
|
||||
retractions)))
|
||||
|
||||
(defn- non-fts-many->queries [ops tx ->SQLite indexing? ref? unique?]
|
||||
(let [q "INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0, index_avet0, index_vaet0, index_fulltext0, unique_value0, sv, svalue_type_tag) VALUES "
|
||||
|
||||
values-part
|
||||
;; e0, a0, v0, tx0, added0, value_type_tag0
|
||||
;; index_avet0, index_vaet0, index_fulltext0,
|
||||
;; unique_value0, sv, svalue_type_tag
|
||||
"(?, ?, ?, ?, 1, ?, ?, ?, 0, ?, ?, ?)"
|
||||
|
||||
repeater (memoize (fn [n] (interpose ", " (repeat n values-part))))]
|
||||
|
||||
;; This query takes ten variables per item. So we partition into max-sql-vars / 10.
|
||||
(map
|
||||
(fn [chunk]
|
||||
(cons
|
||||
;; Query string.
|
||||
(apply str q (repeater (count chunk)))
|
||||
|
||||
;; Bindings.
|
||||
(mapcat (fn [[_ e a v]]
|
||||
(let [[v tag] (->SQLite a v)]
|
||||
[e a v tx tag
|
||||
(indexing? a) ; index_avet
|
||||
(ref? a) ; index_vaet
|
||||
(unique? a) ; unique_value
|
||||
v tag]))
|
||||
chunk)))
|
||||
|
||||
(partition-all (quot max-sql-vars 10) ops))))
|
||||
|
||||
(defn- non-fts-one->queries [ops tx ->SQLite indexing? ref? unique?]
|
||||
(let [first-values-part
|
||||
;; TODO: order value and tag closer together.
|
||||
;; flags0
|
||||
;; sv, svalue_type_tag
|
||||
"(?, ?, ?, ?, ?, ?, ?, ?, 0, ?, ?, ?)"
|
||||
first-repeater (memoize (fn [n] (interpose ", " (repeat n first-values-part))))
|
||||
|
||||
second-values-part
|
||||
"(?, ?, ?, ?, ?, ?)"
|
||||
second-repeater (memoize (fn [n] (interpose ", " (repeat n second-values-part))))
|
||||
]
|
||||
|
||||
;; :db.cardinality/one takes two queries.
|
||||
(mapcat
|
||||
(fn [chunk]
|
||||
[(cons
|
||||
(apply
|
||||
str
|
||||
"INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0, index_avet0, index_vaet0, index_fulltext0, unique_value0, sv, svalue_type_tag) VALUES "
|
||||
(first-repeater (count chunk)))
|
||||
(mapcat (fn [[_ e a v]]
|
||||
(let [[v tag] (->SQLite a v)]
|
||||
[e a v tx 1 tag
|
||||
(indexing? a) ; index_avet
|
||||
(ref? a) ; index_vaet
|
||||
(unique? a) ; unique_value
|
||||
v tag]))
|
||||
chunk))
|
||||
|
||||
(cons
|
||||
(apply
|
||||
str
|
||||
"INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0) VALUES "
|
||||
(second-repeater (count chunk)))
|
||||
(mapcat (fn [[_ e a v]]
|
||||
(let [[v tag] (->SQLite a v)]
|
||||
[e a v tx 0 tag]))
|
||||
chunk))])
|
||||
(partition-all (quot max-sql-vars 11) ops))))
|
||||
|
||||
;;; An FTS insertion happens in two parts.
|
||||
;;; Firstly, we ensure that the fulltext value is present in the store.
|
||||
;;; This is effectively an INSERT OR IGNORE… but FTS tables don't support
|
||||
;;; uniqueness constraints. So we do it through a trigger on a view.
|
||||
;;; When we insert the value, we pass with it a searchid. We'll use this
|
||||
;;; later when inserting the datom.
|
||||
;;; Secondly, we insert a row just like for non-FTS. The only difference
|
||||
;;; is that the value is the rowid into the fulltext_values table.
|
||||
(defn- fts-many->queries [ops tx ->SQLite indexing? ref? unique?]
|
||||
;; TODO: operations with the same text value should be
|
||||
;; coordinated here!
|
||||
;; It'll work fine without so long as queries are executed
|
||||
;; in order and not combined, but even so it's inefficient.
|
||||
(conj
|
||||
(mapcat
|
||||
(fn [[_ e a v] searchid]
|
||||
(let [[v tag] (->SQLite a v)]
|
||||
;; First query: ensure the value exists.
|
||||
[["INSERT INTO fulltext_values_view (text, searchid) VALUES (?, ?)"
|
||||
v searchid]
|
||||
|
||||
;; Second query: tx_lookup.
|
||||
[(str
|
||||
"WITH vv(rowid) AS (SELECT rowid FROM fulltext_values WHERE searchid = ?) "
|
||||
"INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0, index_avet0, index_vaet0, index_fulltext0, unique_value0, sv, svalue_type_tag) VALUES "
|
||||
"(?, ?, (SELECT rowid FROM vv), ?, 1, ?, ?, ?, 1, ?, (SELECT rowid FROM vv), ?)")
|
||||
searchid
|
||||
e a tx tag
|
||||
(indexing? a) ; index_avet
|
||||
(ref? a) ; index_vaet
|
||||
(unique? a) ; unique_value
|
||||
tag]]))
|
||||
ops
|
||||
(range 2000 999999999))
|
||||
["UPDATE fulltext_values SET searchid = NULL WHERE searchid IS NOT NULL"]))
|
||||
|
||||
(defn fts-one->queries [ops tx ->SQLite indexing? ref? unique?]
|
||||
(conj
|
||||
(mapcat
|
||||
(fn [[_ e a v] searchid]
|
||||
(let [[v tag] (->SQLite a v)]
|
||||
;; First query: ensure the value exists.
|
||||
[["INSERT INTO fulltext_values_view (text, searchid) VALUES (?, ?)"
|
||||
v searchid]
|
||||
|
||||
;; Second and third queries: tx_lookup.
|
||||
[(str
|
||||
"WITH vv(rowid) AS (SELECT rowid FROM fulltext_values WHERE searchid = ?) "
|
||||
"INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0, index_avet0, index_vaet0, index_fulltext0, unique_value0, sv, svalue_type_tag) VALUES "
|
||||
"(?, ?, (SELECT rowid FROM vv), ?, 1, ?, ?, ?, 1, ?, (SELECT rowid FROM vv), ?)")
|
||||
searchid
|
||||
e a tx tag
|
||||
(indexing? a) ; index_avet
|
||||
(ref? a) ; index_vaet
|
||||
(unique? a) ; unique_value
|
||||
tag]
|
||||
|
||||
[(str
|
||||
"INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0) VALUES "
|
||||
"(?, ?, (SELECT rowid FROM fulltext_values WHERE searchid = ?), ?, 0, ?)")
|
||||
e a searchid tx tag]]))
|
||||
ops
|
||||
(range 3000 999999999))
|
||||
["UPDATE fulltext_values SET searchid = NULL WHERE searchid IS NOT NULL"]))
|
||||
|
||||
(defn- -run-queries [conn queries exception-message]
|
||||
(go-pair
|
||||
(try
|
||||
(doseq [q queries]
|
||||
(<? (s/execute! conn q)))
|
||||
(catch #?(:clj java.sql.SQLException :cljs js/Error) e
|
||||
(throw (ex-info exception-message {} e))))))
|
||||
|
||||
(defn- -preamble-drop [conn]
|
||||
(let [preamble-drop-index ["DROP INDEX IF EXISTS id_tx_lookup_added"]
|
||||
preamble-delete-tx-lookup ["DELETE FROM tx_lookup"]]
|
||||
(go-pair
|
||||
(p :preamble
|
||||
(doseq [q [preamble-drop-index preamble-delete-tx-lookup]]
|
||||
(<? (s/execute! conn q)))))))
|
||||
|
||||
(defn- -after-drop [conn]
|
||||
(go-pair
|
||||
(doseq [q [;; The lookup table takes space on disk, so we purge it aggressively.
|
||||
["DROP INDEX IF EXISTS id_tx_lookup_added"]
|
||||
["DELETE FROM tx_lookup"]]]
|
||||
(<? (s/execute! conn q)))))
|
||||
|
||||
(defn- -build-transaction [conn tx]
|
||||
(let [build-indices ["CREATE INDEX IF NOT EXISTS idx_tx_lookup_added ON tx_lookup (added0)"]
|
||||
|
||||
;; First is fast, only one table walk: lookup by exact eav.
|
||||
;; Second is slower, but still only one table walk: lookup old value by ea.
|
||||
insert-into-tx-lookup
|
||||
["INSERT INTO tx_lookup
|
||||
SELECT t.e0, t.a0, t.v0, t.tx0, t.added0 + 2, t.value_type_tag0, t.index_avet0, t.index_vaet0, t.index_fulltext0, t.unique_value0, t.sv, t.svalue_type_tag, d.rowid, d.e, d.a, d.v, d.tx, d.value_type_tag
|
||||
FROM tx_lookup AS t
|
||||
LEFT JOIN datoms AS d
|
||||
ON t.e0 = d.e AND
|
||||
t.a0 = d.a AND
|
||||
t.sv = d.v AND
|
||||
t.svalue_type_tag = d.value_type_tag AND
|
||||
t.sv IS NOT NULL
|
||||
|
||||
UNION ALL
|
||||
SELECT t.e0, t.a0, t.v0, t.tx0, t.added0 + 2, t.value_type_tag0, t.index_avet0, t.index_vaet0, t.index_fulltext0, t.unique_value0, t.sv, t.svalue_type_tag, d.rowid, d.e, d.a, d.v, d.tx, d.value_type_tag
|
||||
FROM tx_lookup AS t,
|
||||
datoms AS d
|
||||
WHERE t.sv IS NULL AND
|
||||
t.e0 = d.e AND
|
||||
t.a0 = d.a"]
|
||||
|
||||
t-datoms-not-already-present
|
||||
["INSERT INTO transactions (e, a, v, tx, added, value_type_tag)
|
||||
SELECT e0, a0, v0, ?, 1, value_type_tag0
|
||||
FROM tx_lookup
|
||||
WHERE added0 IS 3 AND e IS NULL" tx] ;; TODO: get rid of magic value 3.
|
||||
|
||||
t-retract-datoms-carefully
|
||||
["INSERT INTO transactions (e, a, v, tx, added, value_type_tag)
|
||||
SELECT e, a, v, ?, 0, value_type_tag
|
||||
FROM tx_lookup
|
||||
WHERE added0 IS 2 AND ((sv IS NOT NULL) OR (sv IS NULL AND v0 IS NOT v)) AND v IS NOT NULL" tx] ;; TODO: get rid of magic value 2.
|
||||
]
|
||||
(go-pair
|
||||
(doseq [q [build-indices insert-into-tx-lookup
|
||||
t-datoms-not-already-present
|
||||
t-retract-datoms-carefully]]
|
||||
(<? (s/execute! conn q))))))
|
||||
|
||||
(defn- -build-datoms [conn tx]
|
||||
(let [d-datoms-not-already-present
|
||||
["INSERT INTO datoms (e, a, v, tx, value_type_tag, index_avet, index_vaet, index_fulltext, unique_value)
|
||||
SELECT e0, a0, v0, ?, value_type_tag0,
|
||||
index_avet0, index_vaet0, index_fulltext0, unique_value0
|
||||
FROM tx_lookup
|
||||
WHERE added0 IS 3 AND e IS NULL" tx] ;; TODO: get rid of magic value 3.
|
||||
|
||||
;; TODO: retract fulltext datoms correctly.
|
||||
d-retract-datoms-carefully
|
||||
["WITH ids AS (SELECT l.rid FROM tx_lookup AS l WHERE l.added0 IS 2 AND ((l.sv IS NOT NULL) OR (l.sv IS NULL AND l.v0 IS NOT l.v)))
|
||||
DELETE FROM datoms WHERE rowid IN ids" ;; TODO: get rid of magic value 2.
|
||||
]]
|
||||
(-run-queries conn [d-datoms-not-already-present d-retract-datoms-carefully]
|
||||
"Transaction violates unique constraint")))
|
||||
|
||||
(defn- -<apply-entities [db tx entities]
|
||||
(let [schema (.-schema db)
|
||||
->SQLite (partial ds/->SQLite schema)
|
||||
fulltext? (memoize (partial ds/fulltext? schema))
|
||||
many? (memoize (fn [a] (ds/multival? schema a)))
|
||||
indexing? (memoize (fn [a] (ds/indexing? schema a)))
|
||||
ref? (memoize (fn [a] (ds/ref? schema a)))
|
||||
unique? (memoize (fn [a] (ds/unique? schema a)))
|
||||
conn (:sqlite-connection db)
|
||||
|
||||
;; Collect all the queries we need to run.
|
||||
queries (atom [])
|
||||
operations (group-by first entities)]
|
||||
|
||||
(when-not (clojure.set/subset? (keys operations) #{:db/retract :db/add})
|
||||
(raise (str "Unknown operations " (keys operations))
|
||||
{:error :transact/syntax, :operations (dissoc operations :db/retract :db/add)}))
|
||||
|
||||
;; We can turn all non-FTS operations into simple SQL queries that we run serially.
|
||||
;; FTS queries require us to get a rowid from the FTS table and use that for
|
||||
;; insertion, so we need another pass.
|
||||
;; We can't just freely use `go-pair` here, because this function is so complicated
|
||||
;; that ClojureScript blows the stack trying to compile it.
|
||||
|
||||
(when-let [retractions (:db/retract operations)]
|
||||
(swap!
|
||||
queries concat (retractions->queries retractions tx fulltext? ->SQLite)))
|
||||
|
||||
;; We want to partition our additions into four groups according to two
|
||||
;; characteristics: whether they require writing to the FTS value table,
|
||||
;; and whether the attribute has a 'many' cardinality constraint. Each of
|
||||
;; these four requires different queries.
|
||||
(let [additions
|
||||
(group-by (fn [[op e a v]]
|
||||
(if (fulltext? a)
|
||||
(if (many? a)
|
||||
:fts-many
|
||||
:fts-one)
|
||||
(if (many? a)
|
||||
:non-fts-many
|
||||
:non-fts-one)))
|
||||
(:db/add operations))
|
||||
transforms
|
||||
{:fts-one fts-one->queries
|
||||
:fts-many fts-many->queries
|
||||
:non-fts-one non-fts-one->queries
|
||||
:non-fts-many non-fts-many->queries}]
|
||||
|
||||
(doseq [[key ops] additions]
|
||||
(when-let [transform (key transforms)]
|
||||
(swap!
|
||||
queries concat
|
||||
(transform ops tx ->SQLite indexing? ref? unique?)))))
|
||||
|
||||
;; Now run each query.
|
||||
;; This code is a little tortured to avoid blowing the compiler stack in cljs.
|
||||
|
||||
(go-pair
|
||||
(<? (-preamble-drop conn))
|
||||
|
||||
(p :run-insert-queries
|
||||
(<? (-run-queries conn @queries "Transaction violates cardinality constraint")))
|
||||
|
||||
;; Follow up by building indices, then do the work.
|
||||
(p :build-and-transaction
|
||||
(<? (-build-transaction conn tx)))
|
||||
|
||||
(p :update-materialized-datoms
|
||||
(<? (-build-datoms conn tx)))
|
||||
|
||||
(<? (-after-drop conn))
|
||||
|
||||
;; Return the written transaction.
|
||||
(p :select-tx-data
|
||||
(mapv (partial row->Datom schema)
|
||||
(<?
|
||||
(s/all-rows
|
||||
(:sqlite-connection db)
|
||||
;; We index on tx, so the following is fast.
|
||||
["SELECT * FROM transactions WHERE tx = ?" tx])))))))
|
||||
|
||||
(defrecord DB [sqlite-connection schema ident-map]
|
||||
;; ident-map maps between keyword idents and integer entids. The set of idents and entids is
|
||||
;; disjoint, so we represent both directions of the mapping in the same map for simplicity. Also
|
||||
|
|
@ -202,18 +523,23 @@
|
|||
(not= 0))))
|
||||
|
||||
(<av [db a v]
|
||||
(let [[v tag] (ds/->SQLite schema a v)]
|
||||
(let [schema (.-schema db) ;; TODO: understand why (schema db) fails.
|
||||
a (entid db a)
|
||||
[v tag] (ds/->SQLite schema a v)
|
||||
yield-datom
|
||||
(fn [rows]
|
||||
(when-let [row (first rows)]
|
||||
(row->Datom schema row)))]
|
||||
(go-pair
|
||||
(->>
|
||||
{:select [:e :a :v :tx [1 :added]] ;; TODO: generalize columns.
|
||||
:from [:all_datoms]
|
||||
:where [:and [:= :index_avet 1] [:= :a a] [:= :value_type_tag tag] [:= :v v]]}
|
||||
(s/format) ;; TODO: format these statements only once.
|
||||
;; TODO: generalize columns.
|
||||
["SELECT e, a, v, tx, 1 AS added FROM all_datoms
|
||||
WHERE index_avet = 1 AND a = ? AND value_type_tag = ? AND v = ?
|
||||
LIMIT 1" a tag v]
|
||||
|
||||
(s/all-rows (:sqlite-connection db))
|
||||
(<?)
|
||||
|
||||
(mapv (partial row->Datom (.-schema db))))))) ;; TODO: understand why (schema db) fails.
|
||||
<?
|
||||
yield-datom))))
|
||||
|
||||
(<next-eid [db tempid]
|
||||
{:pre [(id-literal? tempid)]}
|
||||
|
|
@ -228,161 +554,14 @@
|
|||
{:error :db/bad-part
|
||||
:part (:part tempid)}))
|
||||
|
||||
(<? (exec ["UPDATE parts SET idx = idx + 1 WHERE part = ?" part]))
|
||||
(:eid (first (<? (s/all-rows (:sqlite-connection db) ["SELECT (start + idx) AS eid FROM parts WHERE part = ?" part])))))))
|
||||
(p :next-eid-body
|
||||
(<? (exec ["UPDATE parts SET idx = idx + 1 WHERE part = ?" part]))
|
||||
(:eid (first (<? (s/all-rows (:sqlite-connection db) ["SELECT (start + idx) AS eid FROM parts WHERE part = ?" part]))))))))
|
||||
|
||||
(<apply-entities [db tx entities]
|
||||
{:pre [(db? db) (sequential? entities)]}
|
||||
(go-pair
|
||||
(let [schema (.-schema db)
|
||||
many? (memoize (fn [a] (ds/multival? schema a)))
|
||||
indexing? (memoize (fn [a] (ds/indexing? schema a)))
|
||||
ref? (memoize (fn [a] (ds/ref? schema a)))
|
||||
unique? (memoize (fn [a] (ds/unique? schema a)))
|
||||
<exec (partial s/execute! (:sqlite-connection db))]
|
||||
(p :delete-tx-lookup-before
|
||||
(<? (<exec ["DROP INDEX IF EXISTS id_tx_lookup_added"]))
|
||||
(<? (<exec ["DELETE FROM tx_lookup"])))
|
||||
(-<apply-entities db tx entities))
|
||||
|
||||
(p :insertions
|
||||
(try
|
||||
(doseq [entity entities]
|
||||
(let [[op e a v] entity
|
||||
[v tag] (ds/->SQLite schema a v)
|
||||
fulltext? (ds/fulltext? schema a)]
|
||||
(cond
|
||||
(= op :db/add)
|
||||
(let [v (if fulltext?
|
||||
(<? (<insert-fulltext-value db v))
|
||||
v)]
|
||||
(if (many? a)
|
||||
;; :db.cardinality/many
|
||||
(<? (<exec [(str "INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0, index_avet0, index_vaet0, index_fulltext0, unique_value0, sv, svalue_type_tag) VALUES ("
|
||||
"?, ?, ?, ?, 1, ?, " ; e0, a0, v0, tx0, added0, value_type_tag0
|
||||
"?, ?, ?, ?, " ; flags0
|
||||
"?, ?" ; sv, svalue_type_tag
|
||||
")")
|
||||
e a v tx tag
|
||||
(indexing? a) ; index_avet
|
||||
(ref? a) ; index_vaet
|
||||
fulltext? ; index_fulltext
|
||||
(unique? a) ; unique_value
|
||||
v tag
|
||||
]))
|
||||
;; :db.cardinality/one
|
||||
(do
|
||||
(<? (<exec [(str "INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0, index_avet0, index_vaet0, index_fulltext0, unique_value0, sv, svalue_type_tag) VALUES ("
|
||||
"?, ?, ?, ?, ?, ?, " ; TODO: order value and tag closer together.
|
||||
"?, ?, ?, ?, " ; flags0
|
||||
"?, ?" ; sv, svalue_type_tag
|
||||
")")
|
||||
e a v tx 1 tag
|
||||
(indexing? a) ; index_avet
|
||||
(ref? a) ; index_vaet
|
||||
fulltext? ; index_fulltext
|
||||
(unique? a) ; unique_value
|
||||
v tag
|
||||
]))
|
||||
(<? (<exec [(str "INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0, sv, svalue_type_tag) VALUES ("
|
||||
"?, ?, ?, ?, ?, ?, "
|
||||
"?, ?" ; sv, svalue_type_tag
|
||||
")")
|
||||
e a v tx 0 tag
|
||||
nil nil ; Search values.
|
||||
])))))
|
||||
|
||||
(= op :db/retract)
|
||||
(<? (<exec [(str "INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0, sv, svalue_type_tag) VALUES ("
|
||||
"?, ?, ?, ?, ?, ?, "
|
||||
"?, ?" ; sv, svalue_type_tag
|
||||
")")
|
||||
e a v tx 0 tag
|
||||
v tag
|
||||
]))
|
||||
|
||||
true
|
||||
(raise "Unknown operation at " entity ", expected :db/add, :db/retract"
|
||||
{:error :transact/syntax, :operation op, :tx-data entity}))))
|
||||
|
||||
(catch java.sql.SQLException e
|
||||
(throw (ex-info "Transaction violates cardinality constraint" {} e))))) ;; TODO: say more about the conflicting datoms.
|
||||
|
||||
(p :create-tx-lookup-indices
|
||||
(<? (<exec ["CREATE INDEX IF NOT EXISTS idx_tx_lookup_added ON tx_lookup (added0)"])))
|
||||
|
||||
(p :join
|
||||
(<?
|
||||
(<exec
|
||||
;; First is fast, only one table walk: lookup by exact eav.
|
||||
;; Second is slower, but still only one table walk: lookup old value by ea.
|
||||
["INSERT INTO tx_lookup
|
||||
|
||||
SELECT t.e0, t.a0, t.v0, t.tx0, t.added0 + 2, t.value_type_tag0, t.index_avet0, t.index_vaet0, t.index_fulltext0, t.unique_value0, t.sv, t.svalue_type_tag, d.rowid, d.e, d.a, d.v, d.tx, d.value_type_tag
|
||||
FROM tx_lookup AS t
|
||||
LEFT JOIN datoms AS d
|
||||
ON t.e0 = d.e AND
|
||||
t.a0 = d.a AND
|
||||
t.sv = d.v AND
|
||||
t.svalue_type_tag = d.value_type_tag AND
|
||||
t.sv IS NOT NULL
|
||||
|
||||
UNION ALL
|
||||
SELECT t.e0, t.a0, t.v0, t.tx0, t.added0 + 2, t.value_type_tag0, t.index_avet0, t.index_vaet0, t.index_fulltext0, t.unique_value0, t.sv, t.svalue_type_tag, d.rowid, d.e, d.a, d.v, d.tx, d.value_type_tag
|
||||
FROM tx_lookup AS t,
|
||||
datoms AS d
|
||||
WHERE t.sv IS NULL AND
|
||||
t.e0 = d.e AND
|
||||
t.a0 = d.a
|
||||
"])))
|
||||
|
||||
(p :insert-transaction
|
||||
(p :insert-transaction-added
|
||||
;; Add datoms that aren't already present.
|
||||
(<? (<exec ["INSERT INTO transactions (e, a, v, tx, added, value_type_tag)
|
||||
SELECT e0, a0, v0, ?, 1, value_type_tag0
|
||||
FROM tx_lookup
|
||||
WHERE added0 IS 3 AND e IS NULL" tx]))) ;; TODO: get rid of magic value 3.
|
||||
|
||||
(p :insert-transaction-retracted
|
||||
;; Retract datoms carefully, either when they're matched exactly or when the existing value doesn't match the new value.
|
||||
(<? (<exec ["INSERT INTO transactions (e, a, v, tx, added, value_type_tag)
|
||||
SELECT e, a, v, ?, 0, value_type_tag
|
||||
FROM tx_lookup
|
||||
WHERE added0 IS 2 AND ((sv IS NOT NULL) OR (sv IS NULL AND v0 IS NOT v)) AND v IS NOT NULL" tx])))) ;; TODO: get rid of magic value 2.
|
||||
|
||||
(try
|
||||
(p :update-datoms-materialized-view
|
||||
(p :insert-datoms-added
|
||||
;; Add datoms that aren't already present.
|
||||
(<? (<exec ["INSERT INTO datoms (e, a, v, tx, value_type_tag, index_avet, index_vaet, index_fulltext, unique_value)
|
||||
SELECT e0, a0, v0, ?, value_type_tag0,
|
||||
index_avet0, index_vaet0, index_fulltext0, unique_value0
|
||||
FROM tx_lookup
|
||||
WHERE added0 IS 3 AND e IS NULL" tx])) ;; TODO: get rid of magic value 3.)
|
||||
|
||||
;; TODO: retract fulltext datoms correctly.
|
||||
(p :delete-datoms-retracted
|
||||
(<? (<exec
|
||||
["WITH ids AS (SELECT l.rid FROM tx_lookup AS l WHERE l.added0 IS 2 AND ((l.sv IS NOT NULL) OR (l.sv IS NULL AND l.v0 IS NOT l.v)))
|
||||
DELETE FROM datoms WHERE rowid IN ids" ;; TODO: get rid of magic value 2.
|
||||
])))))
|
||||
|
||||
(catch java.sql.SQLException e
|
||||
(throw (ex-info "Transaction violates unique constraint" {} e)))) ;; TODO: say more about the conflicting datoms.
|
||||
|
||||
;; The lookup table takes space on disk, so we purge it aggressively.
|
||||
(p :delete-tx-lookup-after
|
||||
(<? (<exec ["DROP INDEX IF EXISTS id_tx_lookup_added"]))
|
||||
(<? (<exec ["DELETE FROM tx_lookup"])))
|
||||
|
||||
;; The transaction has been written -- read it back. (We index on tx, so the following is fast.)
|
||||
(let [tx-data (p :select-tx-data
|
||||
(->>
|
||||
(s/all-rows (:sqlite-connection db) ["SELECT * FROM transactions WHERE tx = ?" tx])
|
||||
(<?)
|
||||
|
||||
(mapv (partial row->Datom schema))))]
|
||||
tx-data))))
|
||||
|
||||
(<apply-db-ident-assertions [db added-idents merge]
|
||||
(go-pair
|
||||
|
|
|
|||
|
|
@ -74,7 +74,23 @@
|
|||
;; By default we use Unicode-aware tokenizing (particularly for case folding), but preserve
|
||||
;; diacritics.
|
||||
"CREATE VIRTUAL TABLE fulltext_values
|
||||
USING FTS4 (text NOT NULL, tokenize=unicode61 \"remove_diacritics=0\")"
|
||||
USING FTS4 (text NOT NULL, searchid INT, tokenize=unicode61 \"remove_diacritics=0\")"
|
||||
|
||||
;; This combination of view and triggers allows you to transparently
|
||||
;; update-or-insert into FTS. Just INSERT INTO fulltext_values_view (text, searchid).
|
||||
"CREATE VIEW fulltext_values_view AS SELECT * FROM fulltext_values"
|
||||
"CREATE TRIGGER replace_fulltext_searchid
|
||||
INSTEAD OF INSERT ON fulltext_values_view
|
||||
WHEN EXISTS (SELECT 1 FROM fulltext_values WHERE text = new.text)
|
||||
BEGIN
|
||||
UPDATE fulltext_values SET searchid = new.searchid WHERE text = new.text;
|
||||
END"
|
||||
"CREATE TRIGGER insert_fulltext_searchid
|
||||
INSTEAD OF INSERT ON fulltext_values_view
|
||||
WHEN NOT EXISTS (SELECT 1 FROM fulltext_values WHERE text = new.text)
|
||||
BEGIN
|
||||
INSERT INTO fulltext_values (text, searchid) VALUES (new.text, new.searchid);
|
||||
END"
|
||||
|
||||
;; A view transparently interpolating fulltext indexed values into the datom structure.
|
||||
"CREATE VIEW fulltext_datoms AS
|
||||
|
|
|
|||
|
|
@ -295,7 +295,7 @@
|
|||
(and (id-literal? e)
|
||||
(ds/unique-identity? (db/schema db) a)
|
||||
(not-any? id-literal? [a v]))
|
||||
(let [upserted-eid (:e (first (<? (db/<av db a v))))
|
||||
(let [upserted-eid (:e (<? (db/<av db a v)))
|
||||
allocated-eid (get-in report [:tempids e])]
|
||||
(if (and upserted-eid allocated-eid (not= upserted-eid allocated-eid))
|
||||
(<? (<retry-with-tempid db initial-report initial-entities e upserted-eid)) ;; TODO: not initial report, just the sorted entities here.
|
||||
|
|
|
|||
Loading…
Reference in a new issue