Skip to content

Latest commit

 

History

History
73 lines (64 loc) · 4.38 KB

TODO.md

File metadata and controls

73 lines (64 loc) · 4.38 KB

TODO LIST

TODO

  • (⬜FEAT) allow switching implementations, via hooks and optionally mutable version (an example of this could be to mutate CRDTs via structura)
  • (⬜FEAT) hooks/callbacks for proxy traps or operations
  • (⬜FEAT) add mutable version without copy
  • (🟪DOCS) version in docs
  • (🟪DOCS) write a reference documentation
  • (⬛CODE) try a better solution instead than no-op patch
  • (⬛CODE) enum also for op of json patches
  • (🟦TYPE) better types for json patches
  • (⬜FEAT) maybe use proxy revocable? maybe not necessary because they are already garbage collected
  • (⬛CODE) add more comments in code
  • (⬛CODE) give better names to variables and types
  • (🟫BENC) add benchmarks for patches
  • (🟫BENC) add benchmarks against newer versions of Immer

IN PROGRESS

  • [90%] (🟩TEST) increase test coverage
  • (⬛CODE) correct the nomenclature and behaviour of the helpers
    • original should always return the original object
    • target should be renamed to make more understandable what it does
    • clone should always return the shallow copy of the object or null

SOLVED

  • (🟧BUGS) date methods don't work
  • (🟧BUGS) methods don't work
  • (🟩TEST) type testing
  • (⬜FEAT) json patches don't support "" as key if the path is a string ( to solve this, use a placeholder like empty)
  • (⬜FEAT) "NOTHING" as return, so you can return undefined
  • (⬜FEAT) support async producers,so when you get the result you should check if it is a promise
  • (⬛CODE) split the logic in multiple reusable functions
  • (⬛CODE) organize the code better in multiple files
  • (🟨PERF) turn the proxy trap into a class
  • (🟪DOCS) documentate new settings and remove docs for old helpers, write breaking changes
  • (🟨PERF) store type in the currData
  • (⬜FEAT) full compatibility with this rfc of redux toolkit
  • (🟧BUGS) test fails with both standard patches and auto freeze enabled
  • (🟧BUGS) frozen sets caused infinite loops
  • (🟩TEST) each test should possibly also run on strict copy and/or auto freeze
  • (🟧BUGS) patches don't work with circular references
  • (🟩TEST) generate and try patches in every test
  • (🟨PERF) dummy objects in freeze should be reused as shallow targets
  • (🟨PERF) use object like { [Symbol()]: target } as proxy target
  • (🟧BUGS) reverse patches may create inconsistencies in sets
  • (🟪DOCS) documentate helpers like original and target
  • (⬜FEAT) possibility to convert patches in standard RFC JSON format
  • (⬜FEAT) added support for the "in" operator and ownKeys trap
  • (🟦TYPE) Passing a frozen object should never require explicit casting
  • (🟦TYPE) produceWithPatches didn't allow a return type different from the draft
  • (🟧BUGS) patches could delete an element wrongly with maps
  • (🟧BUGS) assign the descriptor value during strict copy, because the descriptor could be a getter or setter
  • (🟦TYPE) patches should not be callable for primitive types
  • (🟩TEST) add test for symbols in strict mode
  • (🟧BUGS) symbols were not copied on strict copy
  • (🟫BENC) add benchmarks for different setups
  • (⬜FEAT) added patches
  • (🟧BUGS) if I have multiple link to a child for the same parent, the append action is only done once

DISCARDED

  • [ ] (🟨PERF) maybe use only two proxies instead of creating one each time (not doable because you would lose the target ref)
  • [ ] (🟨PERF) freeze the objects while you draft (not doable because if you have any other modifications it will fail them)
  • [ ] (🟨PERF) turn walk parents into a class (actually I tried and it worsen performance)
  • [ ] (⬜FEAT) maybe implement those types? TypedArray, DataView, File, Blob, FileList, DomException
  • [ ] (⬜FEAT) maybe implement those proxy traps? defineProperty, setPrototypeOF, preventExtensions, apply
  • [ ] (🟨PERF) appended: WeakSet could be used to see if an element was external to the tree, so we could avoid cloning it
  • [ ] (🟨PERF) getOwnPropertySymbols adds some overhead on copy: we could disable it with a flag if necessary. Is it worth?
  • [ ] (🟨PERF) typeof v is taken two times, the second time for checking if it's a function. Is it worth solving?