Struct PropertySetJsonbDictionaryBuilder 

pub struct PropertySetJsonbDictionaryBuilder { /* private fields */ }

Custom dictionary builder for PropertySet → JSONB encoding with pointer-based deduplication.

This builder eliminates redundant JSONB serialization and dictionary hash lookups for duplicate PropertySets by using PropertySet’s Arc<Object> pointer addresses as keys.

Performance benefits over Arrow’s BinaryDictionaryBuilder:

• Eliminates content-based hashing: Arrow’s builder hashes JSONB bytes for deduplication
• Pointer-based deduplication: O(1) pointer comparison vs O(n) content hash
• Serialization only when needed: Only serialize PropertySet on first encounter
• Memory efficiency: Shared PropertySet references, single JSONB copy per unique set

Invariant (must not outlive a single parse arena)

The pointer keys are only unique while every appended PropertySet borrows the same parse arena. A dropped arena’s address can be recycled by the next block’s arena, so a builder instance must be fed exactly one block / one arena: construct it fresh in each block processor and finish() it before the arena is dropped. Reusing one builder across arenas would let a recycled address alias a stale entry and emit the wrong JSONB. append_property_set debug-asserts this invariant so any future cross-arena reuse fails loudly in debug/test builds.

Implementations

impl PropertySetJsonbDictionaryBuilder

pub fn new(capacity: usize) -> PropertySetJsonbDictionaryBuilder

Create a new builder with the specified capacity hint

pub fn append_property_set( &mut self, property_set: &PropertySet<'_>, ) -> Result<(), Error>

Append PropertySet using pointer-based deduplication

For cache hits: reuses existing dictionary index (no serialization) For cache misses: serializes once and stores in dictionary

pub fn append_null(&mut self)

Append a null value

pub fn finish(self) -> Result<DictionaryArray<Int32Type>, Error>

Finish building and return the DictionaryArray

Output is identical to Arrow’s BinaryDictionaryBuilder for compatibility

pub fn len(&self) -> usize

Get the current number of appended entries

pub fn is_empty(&self) -> bool

Check if the builder is empty