| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
| |
Now that the bind values are being stored on the actual AST, we need a
way to pull them off into the array that we were previously maintaining
separately. This requires a full walk of the AST. This is an expensive
operation, so I've also added a visitor for delegating to more than one
visitor in a single pass.
|
| |
|
|
|
|
|
|
| |
Switching from `''.dup` to `String.new` had the side effect of changing
the encoding on Ruby 2.4 and later. Oddly, `String.new(encoding:
Encoding::UTF_8)` is significantly slower than `''.dup`. This seems like
a bug in Ruby, but not something we're going to address right now. A
test has been added to ensure this regression doesn't occur again.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
I'm honestly not sure if replacing bind params with their concrete
values is something that belongs in Arel at all, as it's something that
will need to be coupled to the quoting mechanism of the caller, and
could just be accomplished by using `Quoted` instead.
Still, with the new structure we can provide a much simpler API around
substitution. The expectation of the quoter responding to `quote` is a
reasonably minimal API.
I originally used `DelegateClass` here, with the one line override of
`add_bind`, but realized that we have some funky code going on where the
collector returns the next collector to use (in practice `self` is
always returned, and I don't see why we'd ever want to do this).
Removing that would likely be worthwhile, but would be a larger
refactoring
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This is part of a greater refactoring to have the `BindParam` nodes hold
onto their values. We want to generally keep the AST decoupled from what
you're actually doing with those values, but ultimately the usage of
`BindParam` is almost identical to how you'd use `Casted` or `Quoted`.
Forcing consumers of Arel's API to maintain the bind values separately
from the AST makes manipulating the AST essentially impossible, as you
would need to perform a full walk of the AST to determine whether a
given node contains bind parameters, and which value it maps to.
By storing the value on the bind parameter directly, we can collect them
in another AST pass (realistically it'll be part of the same pass that
performs SQL construction for performance reasons). This will
dramatically simplify AST manipulation for Rails or any other consumers
that work with bind params.
As part of this change I've removed the `BindVisitor`, which appears to
be dead code, and had tests break from this change.
|
| |
|
|
|
|
| |
The "bind collector" does not actually collect bind params at all, it
substitutes them out of the final AST, replacing them with a quoted
value.
|
| | |
|
| |
|
|
|
|
| |
This constructor parameter was unused for everything except the
convenience methods `to_sql` and `where_sql`. We can pass the engine
into those methods directly.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
This removes the need for us to do the re-ordering by walking the AST in
ActiveRecord. We're using a block to communicate with the collector,
since the collector needs to be the thing which knows about the index,
while the visitor is the thing that needs to know the syntax. The
BindParam needs to know about neither of these things, so it's been
changed to stop being a subclass of SqlLiteral
I could also see an alternative implementation using format strings if
for some reason blocks cause a problem.
|
| | |
|
| | |
|
| | |
|
| |
|
