perf: avoid mem alloc in formatting, minor bug fix #1521
Conversation
nyurik
changed the title
I don't see how the code could function with that being the case, so it seems possible this has just moved the bug around. I will have to take a closer look. |
Optimize writing formatted string to Serde, bypassing string allocations when possible. Note that `pgrx-sql-entity-graph/src/extension_sql/entity.rs` seemed to have a bug for Function printing - it was printing `Function name)` without the opening parenthesis. Also, in a few spots, I replaced `buffer.write_fmt(format_args!(...))` with `write!(buffer, ...)` -- as that's exactly what `write!` macro is for.
| } | ||
| SqlDeclaredEntity::Function(data) => { | ||
| f.write_str(&(String::from("Function ") + &data.name + ")")) | ||
| write!(f, "Function({})", data.name) |
There was a problem hiding this comment.
I can't actually find a case where this actually diffs a schema, and that is Worrying, because that suggests it is Simply Dead Code.
There was a problem hiding this comment.
I did not exhaustively cover every extension in reach but I have reached the "I need to script out the schema generation delta check" part, which means I might as well bash out a v0.1 solution to #1521
|
any updates on this? |
Optimize writing formatted string to Serde, bypassing string allocations when possible.
Note that
pgrx-sql-entity-graph/src/extension_sql/entity.rsseemed to have a bug for Function printing - it was printingFunction name)without the opening parenthesis.Also, in a few spots, I replaced
buffer.write_fmt(format_args!(...))withwrite!(buffer, ...)-- as that's exactly whatwrite!macro is for.