Improve lookup-ref correctness and performance; reduce in-DB garbage. r=rnewman

Before, the parts of the lookup-ref test that passed took about ~2100ms
on my device.  After, the entire lookup-ref test takes 350ms, a
significant speed-up.  I did not try to plot the graphs of the two
approaches as the number of lookup-refs increased, since the new
approach is significantly better in all cases I can think of.

src/common/datomish/api.cljc

@@ -32,6 +32,8 @@
 
 (def id-literal db/id-literal)
 
+(def lookup-ref db/lookup-ref)
+
 (def db transact/db)
 
 (def entid db/entid)

src/common/datomish/db.cljc

@@ -63,6 +63,23 @@
 (defn id-literal? [x]
   (instance? TempId x))
 
+(defrecord LookupRef [a v])
+
+(defn lookup-ref
+  [a v]
+  (if (and
+        (or (keyword? a)
+            (integer? a))
+        v)
+    (->LookupRef a v)
+    (raise (str "Lookup-ref with bad attribute " a " or value " v
+                {:error :transact/bad-lookup-ref, :a a, :v v}))))
+
+(defn lookup-ref? [x]
+  "Return `x` if `x` is like [:attr value], nil otherwise."
+  (when (instance? LookupRef x)
+    x))
+
 (defprotocol IClock
   (now
     [clock]
@@ -105,6 +122,13 @@
     [db a v]
     "Search for a single matching datom using the AVET index.")
 
+  (<avs
+    [db avs]
+    "Search for many matching datoms using the AVET index.
+
+    Take [[a0 v0] [a1 v1] ...] and return a map {[a0 v0] e0}.  If no datom [e1 a1 v1] exists, the
+    key [a1 v1] is not present in the returned map.")
+
   (<apply-entities
     [db tx entities]
     "Apply entities to the store, returning sequence of datoms transacted.")
@@ -172,9 +196,7 @@
      :table-alias source/gensym-table-alias
      :make-constraints nil}))
 
-;; TODO: make this not do the tx_lookup.  We could achieve this by having additional special values
+(defn- retractAttributes->queries [oeas tx]
-;; of added0, or by separating the tx_lookup table into before and after tables.
-(defn- retractAttributes->queries [eas tx]
   (let [where-part
         "(e = ? AND a = ?)"
 
@@ -183,21 +205,23 @@
       (fn [chunk]
         (cons
           (apply str
-                 "INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0, sv, svalue_type_tag)
+                 "INSERT INTO temp.tx_lookup_after (e0, a0, v0, tx0, added0, value_type_tag0, sv, svalue_type_tag,
-                SELECT e, a, v, ?, 0, value_type_tag, v, value_type_tag
+                  rid, e, a, v, tx, value_type_tag)
+                  SELECT e, a, v, ?, 0, value_type_tag, v, value_type_tag,
+                  rowid, e, a, v, ?, value_type_tag
                   FROM datoms
                   WHERE "
                  (repeater (count chunk)))
+          (cons
+            tx
             (cons
               tx
               (mapcat (fn [[_ e a]]
                         [e a])
-                    chunk))))
+                      chunk)))))
-      (partition-all (quot (dec max-sql-vars) 2) eas))))
+      (partition-all (quot (- max-sql-vars 2) 2) oeas))))
 
-;; TODO: make this not do the tx_lookup.  We could achieve this by having additional special values
+(defn- retractEntities->queries [oes tx]
-;; of added0, or by separating the tx_lookup table into before and after tables.
-(defn- retractEntities->queries [es tx]
   (let [ref-tag (sqlite-schema/->tag :db.type/ref)
 
         ;; TODO: include index_vaet flag here, so we can use that index to speed up the deletion.
@@ -209,27 +233,31 @@
       (fn [chunk]
         (cons
           (apply str
-                 "INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0, sv, svalue_type_tag)
+                 "INSERT INTO temp.tx_lookup_after (e0, a0, v0, tx0, added0, value_type_tag0, sv, svalue_type_tag,
-                SELECT e, a, v, ?, 0, value_type_tag, v, value_type_tag
+                  rid, e, a, v, tx, value_type_tag)
+                  SELECT e, a, v, ?, 0, value_type_tag, v, value_type_tag,
+                  rowid, e, a, v, ?, value_type_tag
                   FROM datoms
                   WHERE "
                  (repeater (count chunk)))
+          (cons
+            tx
             (cons
               tx
               (mapcat (fn [[_ e]]
                         [e e])
-                    chunk))))
+                      chunk)))))
-      (partition-all (quot (dec max-sql-vars) 2) es))))
+      (partition-all (quot (- max-sql-vars 2) 2) oes))))
 
 (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)
+     INSERT INTO temp.tx_lookup_before (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)
+       "INSERT INTO temp.tx_lookup_before (e0, a0, v0, tx0, added0, value_type_tag0, sv, svalue_type_tag)
      VALUES (?, ?, ?, ?, 0, ?, ?, ?)"]
     (map
       (fn [[_ e a v]]
@@ -242,7 +270,7 @@
       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 "
+  (let [q "INSERT INTO temp.tx_lookup_before (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
@@ -290,7 +318,7 @@
         [(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 "
+             "INSERT INTO temp.tx_lookup_before (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)]
@@ -304,7 +332,7 @@
          (cons
            (apply
              str
-             "INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0) VALUES "
+             "INSERT INTO temp.tx_lookup_before (e0, a0, v0, tx0, added0, value_type_tag0) VALUES "
              (second-repeater (count chunk)))
            (mapcat (fn [[_ e a v]]
                      (let [[v tag] (->SQLite a v)]
@@ -341,10 +369,10 @@
           [["INSERT INTO fulltext_values_view (text, searchid) VALUES (?, ?)"
             v searchid]
 
-           ;; Second query: tx_lookup.
+           ;; Second query: 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 "
+              "INSERT INTO temp.tx_lookup_before (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
@@ -365,10 +393,10 @@
           [["INSERT INTO fulltext_values_view (text, searchid) VALUES (?, ?)"
             v searchid]
 
-           ;; Second and third queries: tx_lookup.
+           ;; Second and third queries: 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 "
+              "INSERT INTO temp.tx_lookup_before (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
@@ -378,7 +406,7 @@
             tag]
 
            [(str
-              "INSERT INTO tx_lookup (e0, a0, v0, tx0, added0, value_type_tag0) VALUES "
+              "INSERT INTO temp.tx_lookup_before (e0, a0, v0, tx0, added0, value_type_tag0) VALUES "
               "(?, ?, (SELECT rowid FROM fulltext_values WHERE searchid = ?), ?, 0, ?)")
             e a searchid tx tag]]))
       ops
@@ -390,33 +418,43 @@
     (try
       (doseq [q queries]
         (<? (s/execute! conn q)))
-      (catch #?(:clj java.sql.SQLException :cljs js/Error) e
+      (catch #?(:clj java.lang.Exception :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]]
+       (doseq [q [;; XXX ["DROP INDEX IF EXISTS temp.idx_tx_lookup_before_added"]
-           (<? (s/execute! conn q)))))))
+                  (sqlite-schema/create-temp-tx-lookup-statement "temp.tx_lookup_before")
+                  (sqlite-schema/create-temp-tx-lookup-statement "temp.tx_lookup_after")
+                  ;; TODO: move later, into -build-transaction.
+                  ;; temp goes on index name, not table name.  See http://stackoverflow.com/a/22308016.
+                  (sqlite-schema/create-temp-tx-lookup-eavt-statement "temp.idx_tx_lookup_before_eavt" "tx_lookup_before")
+                  (sqlite-schema/create-temp-tx-lookup-eavt-statement "temp.idx_tx_lookup_after_eavt" "tx_lookup_after")
+                  ["DELETE FROM temp.tx_lookup_before"]
+                  ["DELETE FROM temp.tx_lookup_after"]]]
+         (<? (s/execute! conn q))))))
 
 (defn- -after-drop [conn]
   (go-pair
-    (doseq [q [;; The lookup table takes space on disk, so we purge it aggressively.
+    (p :postamble
-               ["DROP INDEX IF EXISTS id_tx_lookup_added"]
+       ;; TODO: delete tx_lookup_before after filling tx_lookup_after.
-               ["DELETE FROM tx_lookup"]]]
+       (doseq [q [;; XXX ["DROP INDEX IF EXISTS temp.idx_tx_lookup_before_added"]
-      (<? (s/execute! conn q)))))
+                  ["DROP INDEX IF EXISTS temp.idx_tx_lookup_before_eavt"]
+                  ["DROP INDEX IF EXISTS temp.idx_tx_lookup_after_eavt"]
+                  ["DELETE FROM temp.tx_lookup_before"]
+                  ["DELETE FROM temp.tx_lookup_after"]]]
+         (<? (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)"]
+  (let [build-indices ["CREATE INDEX IF NOT EXISTS temp.idx_tx_lookup_added ON tx_lookup_before (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
+        ["INSERT INTO temp.tx_lookup_after
-          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
+          SELECT t.e0, t.a0, t.v0, t.tx0, t.added0, 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
+          FROM temp.tx_lookup_before AS t
           LEFT JOIN datoms AS d
           ON t.e0 = d.e AND
              t.a0 = d.a AND
@@ -425,8 +463,8 @@
              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
+          SELECT t.e0, t.a0, t.v0, t.tx0, t.added0, 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,
+          FROM temp.tx_lookup_before AS t,
                datoms AS d
           WHERE t.sv IS NULL AND
                 t.e0 = d.e AND
@@ -435,14 +473,14 @@
         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
+          FROM temp.tx_lookup_after
-          WHERE added0 IS 3 AND e IS NULL" tx] ;; TODO: get rid of magic value 3.
+          WHERE added0 IS 1 AND e IS NULL" tx]
 
         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
+          FROM temp.tx_lookup_after
-          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.
+          WHERE added0 IS 0 AND ((sv IS NOT NULL) OR (sv IS NULL AND v0 IS NOT v)) AND v IS NOT NULL" tx]
         ]
     (go-pair
       (doseq [q [build-indices insert-into-tx-lookup
@@ -455,13 +493,13 @@
         ["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
+          FROM temp.tx_lookup_after
-          WHERE added0 IS 3 AND e IS NULL" tx] ;; TODO: get rid of magic value 3.
+          WHERE added0 IS 1 AND e IS NULL" tx]
 
         ;; 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)))
+        ["WITH ids AS (SELECT l.rid FROM temp.tx_lookup_after AS l WHERE l.added0 IS 0 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.
+         DELETE FROM datoms WHERE rowid IN ids"
          ]]
     (-run-queries conn [d-datoms-not-already-present d-retract-datoms-carefully]
                   "Transaction violates unique constraint")))
@@ -620,6 +658,61 @@
           <?
           yield-datom))))
 
+  (<avs
+    [db avs]
+    {:pre [(sequential? avs)]}
+
+    (go-pair
+      (let [schema
+            (.-schema db)
+
+            values-part
+            "(?, ?, ?, ?)"
+
+            repeater
+            (memoize (fn [n] (interpose ", " (repeat n values-part))))
+
+            exec
+            (partial s/execute! (:sqlite-connection db))
+
+            ;; Map [a v] -> searchid.
+            av->searchid
+            (into {} (map vector avs (range)))
+
+            ;; Each query takes 4 variables per item. So we partition into max-sql-vars / 4.
+            qs
+            (map
+              (fn [chunk]
+                (cons
+                  ;; Query string.
+                  (apply str "WITH t(searchid, a, v, value_type_tag) AS (VALUES "
+                         (apply str (repeater (count chunk))) ;; TODO: join?
+                         ") SELECT t.searchid, d.e
+                           FROM t, datoms AS d
+                           WHERE d.index_avet IS NOT 0 AND d.a = t.a AND d.value_type_tag = t.value_type_tag AND d.v = t.v")
+
+                  ;; Bindings.
+                  (mapcat (fn [[[a v] searchid]]
+                            (let [a (entid db a)
+                                  [v tag] (ds/->SQLite schema a v)]
+                              [searchid a v tag]))
+                          chunk)))
+
+              (partition-all (quot max-sql-vars 4) av->searchid))
+
+            ;; Map searchid -> e.  There's a generic reduce that takes [pair-chan] lurking in here.
+            searchid->e
+            (loop [coll (transient {})
+                   qs qs]
+              (let [[q & qs] qs]
+                (if q
+                  (let [rs (<? (s/all-rows (:sqlite-connection db) q))
+                        coll* (reduce conj! coll (map (juxt :searchid :e) rs))]
+                    (recur coll* qs))
+                  (persistent! coll))))
+            ]
+        (util/mapvals (partial get searchid->e) av->searchid))))
+
   (<apply-entities [db tx entities]
     {:pre [(db? db) (sequential? entities)]}
     (-<apply-entities db tx entities))

src/common/datomish/db/debug.cljc

@@ -29,6 +29,15 @@
       (filter #(not (= :db/txInstant (second %))))
       (set))))
 
+(defn <datoms>= [db tx]
+  (go-pair
+    (->>
+      (s/all-rows (:sqlite-connection db) ["SELECT e, a, v, tx FROM datoms WHERE tx >= ?" tx])
+      (<?)
+      (mapv #(vector (:e %) (db/ident db (:a %)) (:v %)))
+      (filter #(not (= :db/txInstant (second %))))
+      (set))))
+
 (defn <datoms [db]
   (<datoms-after db 0))
 

src/common/datomish/sqlite_schema.cljc

@@ -26,24 +26,6 @@
    "CREATE UNIQUE INDEX idx_datoms_eavt ON datoms (e, a, value_type_tag, v)"
    "CREATE UNIQUE INDEX idx_datoms_aevt ON datoms (a, e, value_type_tag, v)"
 
-   ;; n.b., v0/value_type_tag0 can be NULL, in which case we look up v from datoms;
-   ;; and the datom columns are NULL into the LEFT JOIN fills them in.
-   ;; TODO: update comment about sv.
-   "CREATE TABLE tx_lookup (e0 INTEGER NOT NULL, a0 SMALLINT NOT NULL, v0 BLOB NOT NULL, tx0 INTEGER NOT NULL, added0 TINYINT NOT NULL,
-                         value_type_tag0 SMALLINT NOT NULL,
-                         index_avet0 TINYINT, index_vaet0 TINYINT,
-                         index_fulltext0 TINYINT,
-                         unique_value0 TINYINT,
-                         sv BLOB,
-                         svalue_type_tag SMALLINT,
-                         rid INTEGER,
-                         e INTEGER, a SMALLINT, v BLOB, tx INTEGER, value_type_tag SMALLINT)"
-
-   ;; Note that `id_tx_lookup_added` is created and dropped
-   ;; after insertion, which makes insertion slightly faster.
-   ;; Prevent overlapping transactions.  TODO: drop added0?
-   "CREATE UNIQUE INDEX idx_tx_lookup_eavt ON tx_lookup (e0, a0, v0, added0, value_type_tag0) WHERE sv IS NOT NULL"
-
    ;; Opt-in index: only if a has :db/index true.
    "CREATE UNIQUE INDEX idx_datoms_avet ON datoms (a, value_type_tag, v, e) WHERE index_avet IS NOT 0"
 
@@ -115,6 +97,35 @@
    "CREATE TABLE parts (part TEXT NOT NULL PRIMARY KEY, start INTEGER NOT NULL, idx INTEGER NOT NULL)"
    ])
 
+(defn create-temp-tx-lookup-statement [table-name]
+  ;; n.b., v0/value_type_tag0 can be NULL, in which case we look up v from datoms;
+  ;; and the datom columns are NULL into the LEFT JOIN fills them in.
+  ;; The table-name is not escaped in any way, in order to allow "temp.dotted" names.
+  ;; TODO: update comment about sv.
+  [(str "CREATE TABLE IF NOT EXISTS " table-name
+        " (e0 INTEGER NOT NULL, a0 SMALLINT NOT NULL, v0 BLOB NOT NULL, tx0 INTEGER NOT NULL, added0 TINYINT NOT NULL,
+           value_type_tag0 SMALLINT NOT NULL,
+           index_avet0 TINYINT, index_vaet0 TINYINT,
+           index_fulltext0 TINYINT,
+           unique_value0 TINYINT,
+           sv BLOB,
+           svalue_type_tag SMALLINT,
+           rid INTEGER,
+           e INTEGER, a SMALLINT, v BLOB, tx INTEGER, value_type_tag SMALLINT)")])
+
+(defn create-temp-tx-lookup-eavt-statement [idx-name table-name]
+  ;; Note that the consuming code creates and drops the indexes
+  ;; manually, which makes insertion slightly faster.
+  ;; This index prevents overlapping transactions.
+  ;; The idx-name and table-name are not escaped in any way, in order
+  ;; to allow "temp.dotted" names.
+  ;; TODO: drop added0?
+  [(str "CREATE UNIQUE INDEX IF NOT EXISTS "
+        idx-name
+        " ON "
+        table-name
+        " (e0, a0, v0, added0, value_type_tag0) WHERE sv IS NOT NULL")])
+
 (defn <create-current-version
   [db]
   (->>

src/common/datomish/transact.cljc

@@ -61,7 +61,7 @@
                      db-after  ;; The DB after the transaction.
                      tx        ;; The tx ID represented by the transaction in this report; refer :db/tx.
                      txInstant ;; The timestamp instant when the the transaction was processed/committed in this report; refer :db/txInstant.
-                     entities  ;; The set of entities (like [:db/add e a v tx]) processed.
+                     entities  ;; The set of entities (like [:db/add e a v]) processed.
                      tx-data   ;; The set of datoms applied to the database, like (Datom. e a v tx added).
                      tempids   ;; The map from id-literal -> numeric entid.
                      part-map  ;; Map {:db.part/user {:start 0x10000 :idx 0x10000}, ...}.
@@ -118,7 +118,7 @@
     true
     entity))
 
-(defn maybe-ident->entid [db [op e a v tx :as orig]]
+(defn maybe-ident->entid [db [op e a v :as orig]]
   ;; We have to handle all ops, including those when a or v are not defined.
   (let [e (db/entid db e)
         a (db/entid db a)
@@ -127,8 +127,8 @@
             (db/entid db v))]
     (when (and a (not (integer? a)))
       (raise "Unknown attribute " a
-             {:form orig :attribute a}))
+             {:form orig :attribute a :entity orig}))
-    [op e a v tx]))
+    [op e a v]))
 
 (defrecord Transaction [db tempids entities])
 
@@ -250,49 +250,43 @@
       ;; Extract the current txInstant for the report.
       (->> (update-txInstant db*)))))
 
-(defn- lookup-ref? [x]
-  "Return `x` if `x` is like [:attr value], false otherwise."
-  (and (sequential? x)
-       (= (count x) 2)
-       (or (keyword? (first x))
-           (integer? (first x)))
-       x))
-
 (defn <resolve-lookup-refs [db report]
   {:pre [(db/db? db) (report? report)]}
 
-  (let [entities (:entities report)]
+  (let [unique-identity? (memoize (partial ds/unique-identity? (db/schema db)))
-    ;; TODO: meta.
+        ;; Map lookup-ref -> entities containing lookup-ref, like {[[:a :v] [[[:a :v] :b :w] ...]] ...}.
+        groups           (group-by (partial keep db/lookup-ref?) (:entities report))
+        ;; Entities with no lookup-ref are grouped under the key (lazy-seq).
+        entities         (get groups (lazy-seq)) ;; No lookup-refs? Pass through.
+        to-resolve       (dissoc groups (lazy-seq)) ;; The ones with lookup-refs.
+        ;; List [[:a :v] ...] to lookup.
+        avs              (set (map (juxt :a :v) (apply concat (keys to-resolve))))
+        ->av             (fn [r] ;; Conditional (juxt :a :v) that passes through nil.
+                           (when r [(:a r) (:v r)]))]
     (go-pair
-      (if (empty? entities)
+      (let [av->e    (<? (db/<avs db avs))
-        report
+            resolve1 (fn [field]
-        (assoc-in
+                       (if-let [[a v] (->av (db/lookup-ref? field))]
-          report [:entities]
+                         (if-not (unique-identity? (db/entid db a))
-          ;; We can't use `for` because go-pair doesn't traverse function boundaries.
+                           (raise "Lookup-ref found with non-unique-identity attribute " a " and value " v
-          ;; Apologies for the tortured nested loop.
+                                  {:error :transact/lookup-ref-with-non-unique-identity-attribute
-          (loop [[op & entity] (first entities)
+                                   :a     a
-                 next (rest entities)
+                                   :v     v})
-                 acc []]
-            (if (nil? op)
-              acc
-              (recur (first next)
-                     (rest next)
-                     (conj acc
-                           (loop [field (first entity)
-                                  rem (rest entity)
-                                  acc [op]]
-                             (if (nil? field)
-                               acc
-                               (recur (first rem)
-                                      (rest rem)
-                                      (conj acc
-                                            (if-let [[a v] (lookup-ref? field)]
                            (or
-                                                ;; The lookup might fail! If so, throw.
+                             (get av->e [a v])
-                                                (:e (<? (db/<av db a v)))
+                             (raise "No entity found for lookup-ref with attribute " a " and value " v
-                                                (raise "No entity found with attr " a " and val " v "."
+                                    {:error :transact/lookup-ref-not-found
-                                                       {:a a :v v}))
+                                     :a     a
-                                              field))))))))))))))
+                                     :v     v})))
+                         field))
+            resolve  (fn [entity]
+                       (mapv resolve1 entity))]
+        (assoc
+          report
+          :entities
+          (concat
+            entities
+            (map resolve (apply concat (vals to-resolve)))))))))
 
 (declare <resolve-id-literals)
 

src/common/datomish/transact/explode.cljc

@@ -47,6 +47,7 @@
       (explode-entity-a-v db entity v straight-a eid)
 
       (and (map? v)
+           (not (db/lookup-ref? v))
            (not (db/id-literal? v)))
       ;; Another entity is given as a nested map.
       (if (ds/ref? (db/schema db) straight-a*)

test/datomish/db_test.cljc

@@ -10,7 +10,7 @@
       [cljs.core.async.macros :as a :refer [go]]))
   (:require
    [datomish.api :as d]
-   [datomish.db.debug :refer [<datoms-after <transactions-after <shallow-entity <fulltext-values]]
+   [datomish.db.debug :refer [<datoms-after <datoms>= <transactions-after <shallow-entity <fulltext-values]]
    [datomish.util :as util #?(:cljs :refer-macros :clj :refer) [raise cond-let]]
    [datomish.schema :as ds]
    [datomish.simple-schema]
@@ -849,5 +849,71 @@
       (is (= (map #(dissoc %1 :db/id) (datomish.simple-schema/simple-schema->schema in))
              expected)))))
 
+(deftest-db test-lookup-refs conn
+  (let [{tx0 :tx} (<? (d/<transact! conn test-schema))
+        {tx1 :tx} (<? (d/<transact! conn [[:db/add 1 :name "Ivan"]
+                                          [:db/add 2 :name "Phil"]
+                                          [:db/add 3 :name "Petr"]]))]
+    (testing "Looks up entity refs"
+      (let [{tx :tx} (<? (d/<transact! conn [[:db/add (d/lookup-ref :name "Ivan") :aka "Devil"]
+                                             [:db/add (d/lookup-ref :name "Phil") :email "@1"]]))]
+        (is (= #{[1 :name "Ivan"]
+                 [2 :name "Phil"]
+                 [3 :name "Petr"]
+                 [1 :aka "Devil"]
+                 [2 :email "@1"]}
+               (<? (<datoms>= (d/db conn) tx1))))))
+
+    (testing "Looks up value refs"
+      (let [{tx :tx} (<? (d/<transact! conn [[:db/add 1 :friends (d/lookup-ref :name "Petr")]
+                                             [:db/add 3 :friends (d/lookup-ref :name "Ivan")]]))]
+        (is (= #{[1 :friends 3]
+                 [3 :friends 1]}
+               (<? (<datoms>= (d/db conn) tx))))))
+
+    (testing "Looks up entity refs in maps"
+      (let [{tx :tx} (<? (d/<transact! conn [{:db/id (d/lookup-ref :name "Phil") :friends 1}]))]
+        (is (= #{[2 :friends 1]}
+               (<? (<datoms>= (d/db conn) tx))))))
+
+    (testing "Looks up value refs in maps"
+      (let [{tx :tx} (<? (d/<transact! conn [{:db/id 2 :friends (d/lookup-ref :name "Petr")}]))]
+        (is (=  #{[2 :friends 3]}
+                (<? (<datoms>= (d/db conn) tx))))))
+
+    (testing "Looks up value refs in sequences in maps"
+      (let [{tx :tx} (<? (d/<transact! conn [{:db/id 1 :friends [(d/lookup-ref :name "Ivan") (d/lookup-ref :name "Phil")]}]))]
+        (is (= #{[1 :friends 1]
+                 [1 :friends 2]}
+               (<? (<datoms>= (d/db conn) tx))))))
+
+    (testing "Looks up refs when there are more than 999 refs (all present)"
+      (let
+          [bound (* 999 2)
+           make-add #(vector :db/add (+ 1000 %) :name (str "Ivan-" %))
+           make-ref #(-> {:db/id (d/lookup-ref :name (str "Ivan-" %)) :email (str "Ivan-" % "@" %)})
+           {tx-data1 :tx-data} (<? (d/<transact! conn (map make-add (range bound))))
+           {tx-data2 :tx-data} (<? (d/<transact! conn (map make-ref (range bound))))]
+        (is (= bound (dec (count tx-data1)))) ;; Each :name is new; dec to account for :db/tx.
+        (is (= bound (dec (count tx-data2)))) ;; Each lookup-ref exists, each :email is new; dec for :db/tx.
+        ))
+
+    (testing "Fails for missing entities"
+      (is (thrown-with-msg?
+            ExceptionInfo #"No entity found for lookup-ref"
+            (<? (d/<transact! conn [[:db/add (d/lookup-ref :name "Mysterioso") :aka "The Magician"]]))))
+      (is (thrown-with-msg?
+            ExceptionInfo #"No entity found for lookup-ref"
+            (<? (d/<transact! conn [[:db/add 1 :friends (d/lookup-ref :name "Mysterioso")]])))))
+
+    (testing "Fails for non-identity attributes"
+      (is (thrown-with-msg?
+            ExceptionInfo #"Lookup-ref found with non-unique-identity attribute"
+            (<? (d/<transact! conn [[:db/add (d/lookup-ref :aka "The Magician") :email "@2"]]))))
+      (is (thrown-with-msg?
+            ExceptionInfo #"Lookup-ref found with non-unique-identity attribute"
+            (<? (d/<transact! conn [[:db/add 1 :friends (d/lookup-ref :aka "The Magician")]])))))))
 
 #_ (time (t/run-tests))
+
+#_ (time (clojure.test/test-vars [#'test-lookup-refs]))