Add a `Once` object to uphold this property. We cannot use the OnceLock
to dictate this, because if we set the OnceLock before actually calling
the callback, there is a moment in time where we will have an incorrect
transient state. This change ensures we atomically call the callback and
then set the OnceLock
Should fix#3217Closes#3217Closes#3237
UNION queries, while useful on their own, are a cornerstone of recursive
CTEs.
This PR implements:
* the merge operator, required to merge both sides of a union query.
* the circuitry necessary to issue the Merge operator.
* extraction of tables mentioned in union and CTE expressions, so we can
correctly populate tables that contain them.
Closes#3234
We currently don't handle non equality, but end up just returning a
bogus result. Let's parse error.
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#3232
The Merge operator is a stateless operator that merges two deltas.
There are two modes: Distinct, where we merge together values that
are the same, and All, where we preserve all values. We use the rowid of
the hashable row to guarantee that: In Distinct mode, the rowid is set
to 0 in both sides. If they values are the same, they will hash to the
same thing. For All, the rowids are different.
The merge operator is used for the UNION statement, which is a
cornerstone of Recursive CTEs.
The population code extracts table information from the select statement
so it can populate the materialized view. But the code, as written
today, is naive. It doesn't capture table information correctly if there
is more than one select statement (such in the case of a union query).
Previously we were rewriting/traversing the AST in a couple different
places, each of these added kinda ad-hoc as we needed them. This
attempts to do the binding of column references as well as the rewriting
of anonymous `Expr::Variable` -> `__param_N` that we use to maintain the
order of bound variables, also normalizes the Qualified Name's.
Also we previously weren't accepting Variable (or at least they wouldn't
work) in places like `LIMIT ? OFFSET ?`, which this PR adds.
I kinda want to keep refactoring translation a bit, and try to break
plan building up into more easy-to-digest chunks.. but I will resist the
urge right now as it's definitely not high priority pre-beta
Reviewed-by: Jussi Saurio <jussi.saurio@gmail.com>
Closes#3210
I don't want to even think about the complexity involved in making sure
that materialized views are still sane after the base table(s) are
altered.
Reviewed-by: Preston Thorpe <preston@turso.tech>
Closes#3223
This PR introduces the final major operator: the JOIN operator.
Many things need to be fixed before we can properly support them, and we
handle those. In particular, JOINs always generate qualified column
statements, but we were not handling them correctly at all in the
operators. Not a problem for linear circuits, but fatal for JOINs.
The operator.rs file also becomes incredibly complex with not one, but
two stateful operator. So it is now broken apart.
Closes#3207
We currently only support column / literal comparisons in the filter
operator. But with JOINs, comparisons are usually against two columns.
Do the work to support it.
Unlike the other operators, project works just fine with ambiguous
columsn, because it works with compiled expressions. We don't need to
patch it, but let's make sure it keeps working by writing a test.
We already did similarly for the AggregateOperator: for joins
you can have the same column name in many tables. And passing schema
information to the operator is a layering violation (the operator may be
operating on the result of a previous node, and at that point there is
no more "schema"). Therefore we pass indexes into the column set the
operator has.
The FilterOperator has a complication: we are using it to generate the
SQL for the populate statement, and that needs column names. However,
we should *not* be using the FilterOperator for that, and that is a
relic from the time where we had operator information directly inside
the IncrementalView.
To enable moving the FilterOperator to index-based, we rework that code.
For joins, we'll need to populate many tables anyway, so we take the
time to do that work here.
This PR improves the DBSP circuit so that it handles the JOIN operator.
The JOIN operator exposes a weakness of our current model: we usually
pass a list of columns between operators, and find the right column by
name when needed.
But with JOINs, many tables can have the same columns. The operators
will then find the wrong column (same name, different table), and
produce incorrect results.
To fix this, we must do two things:
1) Change the Logical Plan. It needs to track table provenance.
2) Fix the aggregators: it needs to operate on indexes, not names.
For the aggregators, note that table provenance is the wrong
abstraction. The aggregator is likely working with a logical table that
is the result of previous nodes in the circuit. So we just need to be
able to tell it which index in the column array it should use.
The operator.rs file was so huge, that we didn't even notice there was a
test block in the middle of the file that was testing things that were
long moved to dbsp.rs (the HashableRow). Move the tests there now.
The join operator is also a stateful operator. It keeps the input deltas
stored in the state, for both the left and right branches of the join.
JOINs extract a join key, which is the values that were used in the
join's equality statement. That key is now our zset_id, and it points
to a collection of rows.
Ahead of the implementation of JOINs, we need to evolve the
IncrementalView, which currently only accepts a single base table,
to keep a list of tables mentioned in the statement.
Our code for view needs to extract the list of columns used in the view.
We currently extract only from "the base table", but once we have joins,
we need a more complex structure, that keeps the mapping of
(tables, columns).
This actually affects both views and materialized views: for views, the
queries with joins work just fine, because views are just aliases for
a query. But the list of columns returned by pragma table_info on the
view is incorrect. We add a test to make sure it is fixed.
For materialized views, we add extensive tests to make sure that the
columns are extracted correctly.
