Commit Graph

3 Commits

Author SHA1 Message Date
Adam Hathcock 715bb7274a Batch by size, Closures and Detached items are computed differently (#164)
* Can debug dependencies

* Different exceptions

* Uses root id only after we found it to signal the end

* DataChunks are created later and need to be accounted for

* format

* use app ids in tests and references

* check sqlite cache after serialize

* use dummy to go through channels to end

* fmt

* Extend channel lib to batch by size

* fmt

* build fix

* adjust limits

* FIx sending

* Optimize reference generation

* more

* remove tolist

* rework closures to be constant and serializer only deals with current....references bases are cached

* fix chunk creation

* another bug fix

* clean up with factories

* add deserializer factory

* Needed to reference interface

* move around streamId

* some clean up

* Use StringBuilder pool on serialization to reduce memory pressure

* remove extra

* remove extra clears

* Fix a flaw in batchsize

* use default complete

* format

* loader should use 1 writer that is batched

* remove redundant ref gen

* Fix graphql commands by adding project id
2024-11-12 14:25:49 +00:00
Adam Hathcock 4cc78c4bc9 Serialize using a Channel (#146)
* Use a stack channel for deserialization

* multi-threaded

* add object dictionary pool

* more pooling

* adjust sqlite transport

* format

* Optimize IsPropNameValid

* object loader first pass

* save test

* add cache pre check

* save better deserialize

* mostly works

* uses tasks but slower at end

* rework to make more sense

* add check to avoid multi-deserialize

* modify max parallelism

* async enqueuing of tasks

* switch to more asyncenumerable

* fmt

* fmt

* cleanup sqlite

* make ServerObjectManager

* revert change

* add ability to skip cache check

* cache json to know what is loaded

* testing

* clean up usage

* clean up and added new op

* Fix exception handling

* fixing progress

* remove codejam

* Hides ObjectPool dependency

* fmt

* Use the 1.0 BCL async to try to be more compatible

* rename to dependencies

* Move Polly to internal dependencies

* format

* remove more old references

* remove stackchannel

* fixes for registration

* remove console writeline

* add cache check shortcut for root object

* start refactoring send

* recevie2 benchmark

* add test for deserialize new

* use channels for sending

* test and fixes

* Use same asyncinterfaces as Dynamo.  Merge fixes

* clean up

* fix download object progress

* put back from bad merge

* intermediate commit: separating get child function from serializer

* send didn't error

* add channels

* Use net48, netstandard2.1 and net8

* remove collection special case

* have to make a tree of tasks even though it may serialize things twice

* pre-id changing during serialize

* need AsyncInterfaces for net48 :(

* options changes

* revert to netstandard2.0 and net8.0

* fix totals

* revert httpcontext changes

* format

* clean up

* active tasks works when accounting for id not being stable

* add id tests

* more fixes

* works

* format

* Convert to BaseItem and use single SQLite checks to avoid locks

* use locks and batch sqlite operations

* hook up and handle null ids

* remove unused parameter

* remove progress from serializer itself

* invert has objects call

* readd object references

* format

* fix tests

* remove active tasks check

* bug fix for json cache

* remove locks from sqlite

* General Send test

* add childclosures

* redo extract all to be enumerable

* group tests in projects

* caching json does matter

* cache checking should be managed by channels

* format

* Merge pull request #152 from specklesystems/new-json-test

Uses a new objects test in Revit for serialization tests

* add skip

* add new roundtrip test

* fix finish

* clean up tests

* check happens in serialize...don't do it twice

* better progress reporting

* fix progress reporting

* only use detached properties when children gathering

* move detached tests

* add detached tests

* fix merge

* Fix progress change

* fix more tests

---------

Co-authored-by: Jedd Morgan <45512892+JR-Morgan@users.noreply.github.com>
Co-authored-by: Claire Kuang <kuang.claire@gmail.com>
2024-11-05 09:56:54 +00:00
Adam Hathcock cca8828565 Using Tasks for Deserialization (#143)
* Use a stack channel for deserialization

* multi-threaded

* add object dictionary pool

* more pooling

* adjust sqlite transport

* format

* Optimize IsPropNameValid

* object loader first pass

* save test

* add cache pre check

* save better deserialize

* mostly works

* uses tasks but slower at end

* rework to make more sense

* add check to avoid multi-deserialize

* modify max parallelism

* async enqueuing of tasks

* switch to more asyncenumerable

* fmt

* fmt

* cleanup sqlite

* make ServerObjectManager

* revert change

* add ability to skip cache check

* cache json to know what is loaded

* testing

* clean up usage

* clean up and added new op

* Fix exception handling

* fixing progress

* remove codejam

* remove stackchannel

* remove console writeline

* add cache check shortcut for root object

* recevie2 benchmark

---------

Co-authored-by: Jedd Morgan <45512892+JR-Morgan@users.noreply.github.com>
2024-10-22 14:14:39 +01:00