* Pre: remove remnants of 'open_empty'
* Pre: Cleanup 'datoms' table after a timeline move
Since timeline move operations use a transactor, they generate a
"phantom" 'tx' and a 'txInstant' assertion. It is "phantom" in a sense
that it was never present in the 'transactions' table, and is entirely
synthetic as far as our database is concerned.
It's an implementational artifact, and we were not cleaning it up.
It becomes a problem when we start inserting transactions after a move.
Once the transactor clashes with the phantom 'tx', it will retract the
phantom 'txInstant' value, leaving the transactions log in an incorrect state.
This patch adds a test for this scenario and elects the easy way out: simply
remove the offending 'txInstant' datom.
* Part 1: Sync without support for side-effects
A "side-effect" is defined here as a mutation of a remote state as part
of the sync.
If, during a sync we determine that a remote state needs to be changed, bail out.
This generally supports different variations of "baton-passing" syncing, where clients
will succeed syncing if each change is non-conflicting.
* Part 2: Support basic "side-effects" syncing
This patch introduces a concept of a follow-up sync. If a sync generated
a "merge transaction" (a regular transaction that contains assertions
necessary for local and remote transaction logs to converge), then
this transaction needs to be uploaded in a follow-up sync.
Generated SyncReport indicates if a follow-up sync is required.
Follow-up sync itself is just a regular sync. If remote state did not change,
it will result in a simple RemoteFastForward. Otherwise, we'll continue
merging and requesting a follow-up.
Schema alterations are explicitly not supported.
As local transactions are rebased on top of remote, following changes happen:
- entids are changed into tempids, letting transactor upsert :db/unique values
- entids for retractions are changed into lookup-refs if we're confident they'll succeed
-- otherwise, retractions are dropped on the floor
* Post: use a macro for more readable tests
* Tolstoy README
Sync needs to operate over a "mentat transaction", not just a "db transaction".
This shuffle allows internal mentat crates to consume InProgress, which models
the concept of a "mentat transaction".
* Add a top-level "syncable" feature.
Tested with:
cargo test --all
cargo test --all --no-default-features
cargo build --manifest-path tools/cli/Cargo.toml --no-default-features
cargo run --manifest-path tools/cli/Cargo.toml --no-default-features debugcli
Co-authored-by: Nick Alexander <nalexander@mozilla.com>
* Add 'syncable' feature to 'db' crate to conditionally derive serialization for Partition*
This is leading up to syncing with partition support.
I elected to keep Tolstoy using `failure::Error`, because Tolstoy
looks rather more like a high-level application (and will continue to
do so for a while) than a production-ready mid- or low-level API.
This uses a `SELECT *` from an inner subselect to filter potentially `NULL` aggregates.
The alternative is to handle `NULL` values throughout the projector, which is simple but loses a valuable invariant: Mentat SQL queries produce values that are not `NULL`.
This should address #663, by re-inserting type checking in the
transactor stack after the entry point used by the term builder.
Before this commit, we were using an SQLite UNIQUE index to assert
that no `[e a]` pair, with `a` a cardinality one attribute, was
asserted more than once. However, that's not in line with Datomic,
which treats transaction inputs as a set and allows a single datom
like `[e a v]` to appear multiple times. It's both awkward and not
particularly efficient to look for _distinct_ repetitions in SQL, so
we accept some runtime cost in order to check for repetitions in the
transactor. This will allow us to address #532, which is really about
whether we treat inputs as sets. A side benefit is that we can
provide more helpful error messages when the transactor does detect
that the input truly violates the cardinality constraints of the
schema.
This commit builds a trie while error checking and collecting final
terms, which should be fairly efficient. It also allows a simpler
expression of input-provided :db/txInstant datoms, which in turn
uncovered a small issue with the transaction watcher, where-by the
watcher would not see non-input-provided :db/txInstant datoms.
This transition to Datomic-like input-as-set semantics allows us to
address #532. Previously, two tempids that upserted to the same entid
would produce duplicate datoms, and that would have been rejected by
the transactor -- correctly, since we did not allow duplicate datoms
under the input-as-list semantics. With input-as-set semantics,
duplicate datoms are allowed; and that means that we must allow
tempids to be equivalent, i.e., to resolve to the same tempid.
To achieve this, we:
- index the set of tempids
- identify tempid indices that share an upsert
- map tempids to a dense set of contiguous integer labels
We use the well-known union-find algorithm, as implemented by
petgraph, to efficiently manage the set of equivalent tempids.
Along the way, I've fixed and added tests for two small errors in the
transactor. First, don't drop datoms resolved by upsert (#679).
Second, ensure that complex upserts are allocated.
I don't know quite what happened here. The Clojure implementation
correctly kept complex upserts that hadn't resolved as complex
upserts (see
9a9dfb502a/src/common/datomish/transact.cljc (L436))
and then allocated complex upserts if they didn't resolve (see
9a9dfb502a/src/common/datomish/transact.cljc (L509)).
Based on the code comments, I think the Rust implementation must have
incorrectly tried to optimize by handling all complex upserts in at
most a single generation of evolution, and that's just not correct.
We're effectively implementing a topological sort, using very specific
domain knowledge, and its not true that a node in a topological sort
can be considered only once!
* Make properties on NamespacedKeyword/NamespacedSymbol private
* Use only a single String for NamespacedKeyword/NamespacedSymbol
* Review comments.
* Remove unsafe code in namespaced_name.
Benchmarking shows approximately zero change.
* Allow the types of ns and name to differ when constructing a NamespacedName.
* Make symbol namespaces optional.
* Normalize names of keyword/symbol constructors.
This will make the subsequent refactor much less painful.
* Use expect not unwrap.
* Merge Keyword and NamespacedKeyword.
There are few reasons to do this:
- it's difficult to add symbol interning to combine-based parsers like
tx-parser -- literally every type changes to reflect the interner,
and that means every convenience macro we've built needs to chagne.
It's trivial to add interning to rust-peg-based parsers.
- combine has rolled forward to 3.2, and I spent a similar amount of
time investigating how to upgrade tx-parser (to take advantage of
the new parser! macros in combine that I think are necessary for
adapting to changing types) as I did just converting to rust-peg.
- it's easy to improve the error messages in rust-peg, where-as I have
tried twice to improve the nested error messages in combine and am
stumped.
- it's roughly 4x faster to parse strings directly as opposed to
edn::ValueAndSpan, and it'll be even better when we intern directly.
* Refactor AttributeCache populator code for use from pull.
* Pre: add to_value_rc to Cloned.
* Pre: add From<StructuredMap> for Binding.
* Pre: clarify Store::open_empty.
* Pre: StructuredMap cleanup.
* Pre: clean up a doc test.
* Split projector crate. Pass schema to projector.
* CLI support for printing bindings.
* Add and use ConjoiningClauses::derive_types_from_find_spec.
* Define pull types.
* Implement pull on top of the attribute cache layer.
* Add pull support to the projector.
* Parse pull expressions.
* Add simple pull support to connection objects.
* Tests for pull.
* Compile with Rust 1.25.
The only choice involved in this commit is that of replacing the
anonymous lifetime '_ with a named lifetime for the cache; since we're
accepting a Known, which includes the cache in question, I think it's
clear that we expect the function to apply to any given cache
lifetime.
* Review comments.
* Bail on unnamed attribute.
* Make assert_parse_failure_contains safe to use.
* Rework query parser to report better errors for pull.
* Test for mixed wildcard and simple attribute.
* Pre: eliminate some occurrences of Rc, largely through the magic of Into.
* Pre: introduce FromRc to convert between refcounted types.
* Introduce ValueRc as an abstraction over Rc/Arc choice.
* Move Cloned to core.
* Move CString-creation methods to TypedValue.
* Finish transition.
* Pre: clean up core/src/lib.rs.
* Pre: use indexmap 1.0 in db and query-projector.
* Change rel results to be a RelResult instance, not a Vec<Vec<TypedValue>>.
This avoids memory fragmentation and improves locality by using a single
heap-allocated vector for all bindings, rather than a separate
heap-allocated vector for each row.
We hide this abstraction behind the `RelResult` type, which tracks the
stride length (width) of each row.
* Don't allocate temporary vectors when projecting RelResults.
`tx-ids` allows to enumerate transaction IDs efficiently.
`tx-data` allows to extract transaction log data efficiently.
We might eventually allow to filter by impacted attribute sets as well.
* Pre: use debugcli in VSCode.
* Pre: wrap subqueries in parentheses in output SQL.
* Pre: add ExistingColumn.
This lets us make reference to columns by name, rather than only
pointing to qualified aliases.
* Pre: add Into for &str to TypedValue.
* Pre: add Store.transact.
* Pre: cleanup.
* Parse and algebrize simple aggregates. (#312)
* Follow-up: print aggregate columns more neatly in the CLI.
* Useful ValueTypeSet helpers.
* Allow for entity inequalities.
* Add 'differ', which is a ref-specialized not-equals.
* Add 'unpermute', a function for getting unique, distinct pairs from bindings.
* Review comments.
* Add 'the' pseudo-aggregation operator.
This allows for a corresponding value to be returned when a query
includes one 'min' or 'max' aggregate.
* Use the cache to make constant queries super fast.
* Fix translate tests to match: we no longer generate SQL for many of them!
* Accumulate additions and removals into the cache.
* Make attribute cache clone-on-write; store it in Metadata.
* Allow caching of fulltext attributes, interning strings.
This puts caching in mentat_db, adds a reverse lookup capability for
unique attributes, and populates bidirectional caches with a single
SQL cursor walk.
Differentiate between begin_read and begin_uncached_read.
Note that we still allow toggling within InProgress, because there might be
transient local state that makes starting a new transaction impossible.
* Define Store, which is a simple container for a SQLite connection and a Conn.
This is a breaking change.
* Return the FindSpec as part of QueryOutput, not just results.
* Switch to using stderr in appropriate places in CLI.
* Print columns in CLI output.
* Add TypedValue::instant(micros).
* Add From<f64> for TypedValue.
* Add lookup_values_for_attribute to Conn.
* Add q_explain to Queryable.
* Expose an iterator over FindSpec's columns.
* Export edn from mentat crate. Export QueryExecutionResult.
* Implement Display for Variable and Element.
* Introduce a `kw` macro.
This allows you to write:
```rust
kw!(:foo/bar)
```
instead of
```rust
NamespacedKeyword::new("foo", "bar")
```
… and it's more efficient, too.
Add `mentat::open`, eliminate use of `mentat_db` in some places.
Gregory Burd
