Match or exceed v1 performance#580

Open

jhollinger wants to merge 50 commits into

release-2.0from

jh/release-2.0-faster

Commits on May 21, 2026

V2 serializer refactor for huge perf improvement
jhollinger
committed
Have dedicated structs for field reflection so people don't depend on internal details
jhollinger
committed
Add options to reflections
jhollinger
committed
Cleanup
jhollinger
committed
Reduce V2 serializer allocations by ~23% via fast path

scottmyron
authored and
jhollinger
committed
Get back to a single extract & serialization path. Sacrifices a *little* speed but should be easier to test & maintain
jhollinger
committed
Ensure field conditional & default status is checked after copying down blueprint options
jhollinger
committed
Too difficult to use separate internal/reflected field defs. Use internal but prefix private fields with an "_"
jhollinger
committed
Allow options to be passed when multiple fields are defined with `fields`
jhollinger
committed
Centralize incrementing serialization depth
jhollinger
committed
Test to prove there's a way to share partials across classes
jhollinger
committed
Replace instance_exec with simple calls in most cases.
jhollinger
committed
Simplify extraction code
jhollinger
committed
Rename Context::Render to Context::Init
jhollinger
committed
Only allow around_result to change options. And don't freeze them until after that hook runs
jhollinger
committed
Since V2 doesn't currently support init-per-render extensions, it doesn't make sense to allow classes or procs
jhollinger
committed
Move ExtensionHelpers back into Extension since it's only used in one place now
jhollinger
committed
More granular decisions on allocating field context objects (will save time/memory)
jhollinger
committed
Rename the 'from' option to 'source'. Reads much better in the docs
jhollinger
committed
Allow blueprint options and most field attributes to be modified by around_blueprint_init. Use a single field type for fields, objects, and collections.
jhollinger
committed
Add back the around_blueprint hook (with optimized context obj allocation for collections)
jhollinger
committed
Optimization for around_blueprint_init when nothing but field order or inclusion is changed
jhollinger
committed
Put all the field classes under a module
jhollinger
committed
'default' Procs/methods don't need the value. For 'default_if', pass value as 2nd arg (so that ctx is always first arg)
jhollinger
committed
Rename the core Wrapper extension to Root.
jhollinger
committed
Extensions to make if, unless, and default_if V1-compatible
jhollinger
committed
Extension to make V1's 'name' option work
jhollinger
committed
Extension to make V1's 'extractor' option work
jhollinger
committed
Don't create field contexts if the block doesn't use them
jhollinger
committed
Rubocop and cleanup
jhollinger
committed
Alias 'excludes' to 'exclude' for backwards compatibility
jhollinger
committed

Commits on May 27, 2026

Commits on Jun 2, 2026

Commits on Jun 3, 2026

Commits on Jun 5, 2026

Commits on Jun 8, 2026

Allow excluding formatters. Use 'exclude' with options instead of dedicated methods
jhollinger
committed

Commits on Jun 9, 2026

Commits on Jun 13, 2026

In the DSL, separate excluding inherited elements from excluding elements from partials.
jhollinger
committed

Commits on Jun 15, 2026

Represent the fully evaled Blueprint as a struct, instead of as attrs on the Blueprint
jhollinger
committed