Commit graph

41 commits

Author SHA1 Message Date
Nick Alexander
63b304ea5f Start importing places. This is just about profiling for now. 2016-08-19 12:40:12 -07:00
Nick Alexander
badec36aaa Completely rewrite main transaction logic to be faster.
This is almost complete; it passes the test suite save for retracting
fulltext datoms correctly.

There's a lot to say about this approach, but I don't have time to give
too many details.  The broad outline is as follows.  We collect datoms
to add and retract in a tx_lookup table.  Depending on flags ("search
value" sv and "search value type tag" svalue_type_tag) we "complete" the
tx_lookup table by joining matching datoms.  This allows us to find
datoms that are present (and should not be added as part of the
transaction, or should be retracted as part of the transaction, or
should be replaced as part of the transaction.  We complete the
tx_lookup (in place!) in two separate INSERTs to avoid a quadratic
two-table walk (explain the queries to observe that both INSERTs walk
the lookup table once and then use the datoms indexes to complete the
matching values).

We could simplify the code by using multiple lookup tables, both for the
two cases of search parameters (eav vs. ea) and for the incomplete and
completed rows.  Right now we differentiate the former with NULL checks,
and the latter by incrementing the added0 column.  It performs well
enough, so I haven't tried to understand the performance of separating
these things.

After the tx_lookup table is completed, we build the transaction from
it; and update the datoms materialized view table as well.  Observe the
careful handling of the "search value" sv parameters to handle replacing
:db.cardinality/one datoms.

Finally, we read the processed transaction back to produce to the API.
This is strictly to match the Datomic API; we might make allow to skip
this, since many consumers will not want to stream this over the wire.

Rough timings show the transactor processing a single >50k datom
transaction in about 3.5s, of which less than 0.5s is spent in the
expensive joins.  Further, repeating the processing of the same
transaction is only about 3.5s again!  That's the worst possible for the
joins, since every single inserted datom will already be present in the
database, making the most expensive join match every row.
2016-08-19 12:40:11 -07:00
Nick Alexander
4a46bdd1bd Extract datomish.db.debug from test code, in order to use it during development. 2016-08-19 12:27:40 -07:00
Nick Alexander
9aed24ae39 Fix :db.unique/value, which should be per (a, v) pair, not per v-value. 2016-08-19 12:27:40 -07:00
Nick Alexander
3dfdea99e7 Implement parts: Make the DB allocate and persist entity IDs.
This implementation is inefficient because each allocated temporary ID
touches the database, but it's enough to allow to re-open DBs.
2016-08-19 12:27:39 -07:00
Nick Alexander
29d409be64 Tag values with value type tags in SQLite. 2016-08-19 12:27:39 -07:00
Nick Alexander
b4e5c88d6a Add d/q; make query minimally schema aware. 2016-08-19 12:27:39 -07:00
Nick Alexander
cfe753a3bd Add d/{ident,entid} for mapping between keyword idents and integer entids. 2016-08-19 12:27:39 -07:00
Richard Newman
d687d4221a Clarify d/<connect. 2016-08-19 12:27:39 -07:00
Richard Newman
1a54cec1ba Use correct view for fulltext+non-fulltext. 2016-08-05 16:29:21 -07:00
Richard Newman
fc845a9950 Implement basic fulltext binding. r=nalexander 2016-08-05 13:04:06 -07:00
Richard Newman
4c2230929e Turn Source into a protocol. Allow source->from to switch on attribute. r=nalexander 2016-08-05 13:03:48 -07:00
Nick Alexander
c948902c52 Follow-up: Extract datomish.transact.explode.
This required pushing the ID literal out of transact.  I elected to put
them in DB, since literal allocation will end up in IDB eventually.
2016-08-04 17:54:50 -07:00
Nick Alexander
d8c976c3ad Follow-up: split the monolith!
This was a little more tricky than might be expected because the
initialization process uses the transactor to bootstrap the database.
Since Clojure doesn't accept mutually recursive modules, this
necessitated a third module, namely "db-factory", which uses both "db"
and "transact".  While I was here, I started an "api" module, to paper
over the potentially complicated internal module structure for external
consumers.  In time, this "api" module may also grow CLJS-specific JS
transformations.
2016-08-04 17:54:32 -07:00
Nick Alexander
73b155cfdc Follow-up: don't accept schema when creating DB.
This schema was already ignored.
2016-08-04 16:34:48 -07:00
Nick Alexander
44db8116bf Handle nested maps, sequences in maps, and reverse references. 2016-08-04 14:26:20 -07:00
Nick Alexander
d9a8cb0d6a Don't accept user-provided tx values.
This agrees with Datomic.  DataScript allows tx values, possibly to
allow reconstructing DBs from Datom streams, but appears to handle
user-provided tx values in the transactor inconsistently.
2016-08-04 14:26:20 -07:00
Nick Alexander
38545f6efc Add :tx and :txInstant to TxReport; accept :db/tx in transactor; allow to set :db/txInstant.
The implementation of :db/tx is special and may need to change over
time.  We add it as a special ident, with value the current transaction
entity ID, specified per-transaction.  This works well right now but
introduces some (internal) ordering requirements that may need to be
loosened.
2016-08-04 14:26:20 -07:00
Nick Alexander
f25838a1eb Support :db/fulltext true.
Internally, we use SQLite's FTS4 to maintain a fulltext_values table of
unique "text" values.  Fulltext indexed datoms have value v that is the
rowid into fulltext_values.  We manually maintain the map between rowid
and value in the transactor.

For convenience, we expose two views interpolating the real text values
into the datoms structure.
2016-08-04 14:26:20 -07:00
Nick Alexander
13f33a4915 Unify test pattern around "(... -after tx0)". 2016-08-04 14:26:20 -07:00
Nick Alexander
296c9cb436 Bootstrap DB schema; persist and restore schema from materialized views. 2016-08-04 14:26:20 -07:00
Nick Alexander
6a8739bd2f Accept new schema fragments via :db.part/db :db.install/attribute. 2016-08-04 14:26:20 -07:00
Nick Alexander
9497d69b44 Respect :db/unique constraints; test upserts.
This version includes SQLite-level unique indexes; these should never be
needed.  I've included them as a fail-safe while testing; they'll help
us catch errors in the transaction layer above.
2016-08-04 14:26:20 -07:00
Nick Alexander
80742242e2 Fix cljs 2016-08-04 14:26:20 -07:00
Nick Alexander
7a90c43a5a Map valueTypes to SQLite encodings.
In the future, we might add a layer of indirection, hashing values to
avoid duplicating storage, or sorting URLs, or handling fulltext indexed
values differently, or ...
2016-08-04 14:26:20 -07:00
Nick Alexander
43423b7d0a Validate value types. 2016-08-04 14:26:20 -07:00
Nick Alexander
661e7ed123 Allow to add new :db/ident mappings. 2016-08-04 14:26:20 -07:00
Nick Alexander
baec3815b0 Implement transactions. 2016-08-04 14:26:20 -07:00
Richard Newman
8a77dcd8f0 Implement simple 'or' clauses. r=nalexander 2016-07-27 17:04:32 -07:00
Richard Newman
1ad67a03eb Add tests and comments for clause ordering. 2016-07-26 11:19:51 -07:00
Richard Newman
feaca75a74 Add tests for joins and not-clauses. 2016-07-26 10:51:13 -07:00
Richard Newman
72f7c8e1a3 Fix test-raise. 2016-07-26 10:50:55 -07:00
Richard Newman
42361c1e5e Implement negation, predicates, external scalar bindings, <?q. r=nalexander 2016-07-25 17:08:00 -07:00
Nick Alexander
13dc1cc9b6 Take more general raise from DataScript. 2016-07-19 10:35:35 -07:00
Nick Alexander
bf080ced3c Part 2: Add ISQLiteConnectionFactory and CLJC test that uses it. 2016-07-13 18:19:22 -07:00
Nick Alexander
cca5010671 Part 1: Fix testing errors.
Some of these were just typos, but `with-open` was fatally flawed on
CLJS (we couldn't call `.close` at all), and `deftest-async` was hiding
all failures (due to a typo).
2016-07-13 18:19:22 -07:00
Nick Alexander
900e77862e Implement database on top of SQLite connection. r=rnewman
We would prefer to talk about a knowledge base on top of a database, but
all the Datomic and DataScript code (and symbols, like :db/add, etc)
refer to the "database of datoms", so let's roll with that nomenclature
and try to be specific that the persistent storage-layer is SQLite.
This will become more clear when we actually use SQLite's unique
capabilities for text indexing.
2016-07-13 12:37:23 -07:00
Richard Newman
db68a714f6 First pass at translating Datalog queries into SQL. r=nalexander
Signed-off-by: Richard Newman <rnewman@twinql.com>
2016-07-13 09:51:42 -07:00
Nick Alexander
724c37466d Add an SQLite connection abstraction. 2016-07-12 13:56:26 -07:00
Nick Alexander
0a312b4f40 Add an async and async testing framework.
This is a well-worn idea: use a `promise-channel` of `[result nil]` or
`[nil error]` pairs.  The `go-pair` and `<?` macros handle catching
exceptions (important, given that synchronous CLJ code expects to throw
rather than return an error promise or similar), allowing code like:
```
(go-pair
  (let [result (<? (pair-chan-fn))]
    (when (not result)
      (throw (Exception. "No result!")))
    (transform result)))
```
to be expressed naturally.  These are the equivalents of `async` and
`await` in JS.

The implementation is complicated by significant incompatibilities
between CLJ and CLJS.  The solution presented here takes care to
separate the macro definitions into CLJ.  Sadly, this requires
namespacing the per-environment symbols explicitly; but we hope to
minimize such code in files like this.

The most significant restriction to this approach is that consumers must
require the transitive dependencies of the macro-defining modules.  See
the included tests (both CLJ and CLJS) for the appropriate
incantations (for pair-chan, core.async, and test).
2016-07-12 13:56:26 -07:00
Nick Alexander
08f7084d92 Add lein-doo to run ClojureScript tests. 2016-07-06 16:42:40 -07:00