CHANGELOG.md

Changelog

0.11.2

Bug Fixes

• Fix none lookup result for optional pointers to accessory components

New Features

• Improve readability of EntityDebug output

0.11.1

Bug Fixes

• Resolved a crash when looking up a pointer for an optional accessory
• Fix invalid dumped code for temporary variables and queries over accessories

New Features

• Runtime performance optimizations

0.11.0

Bug Fixes

• Fix inbox sharing when a system is assigned to a variable
• Fix inbox name collisions when argument names were the same
• Fix error messages when a system incorrectly returns a value
• Fix invalid code dumping when a newline is injected after a sink parameter
• Allow joined tuples to operated on joined tuples

Backwards incompatible changes

• Use () operator for directives instead of passing in procs. This improves compile speed and performance by removing the need for Nim to create and manage closures. However, it requires that all Necsus users enable the --experimental:callOperator flag.
• Remove the necsusFloat32 flag and instead use BiggestUInt and BiggestFloat

New Features

• Accessory components
  • Adding the accessory pragma means that a component no longer forces the creation of new archetypes. This reduces the size of generated code and improves compile speed.
• Adds SystemVar.clear
• Support tuple joining without explicit type definitions
• Add an error message when attach/detach/swap fails
• Build speed improvements
• Adds SaveSystemInstance and EventSystemInstance types
• Fast compile mode for speeding up IDE integration

0.10.0

Bug Fixes

• Fix memory corruption bug caused by Nim mishandling sink parameters
• Support tuples in Shared and Local
• Fix a bug where an instanced eventSys can't be invoked

New Features

• Add compiler flags for tracing various behavior during execution
  • -d:necsusSaveTrace -- Log save and restore activity
  • -d:necsusQueryTrace -- Log executed queries
  • -d:necsusEventTrace -- Log when an event is sent
  • -d:necsusEntityTrace -- Log when entities are created, modified or deleted
• Overall reduction of memory allocations

Backwards incompatible changes

• Mark the logger parameter as gcsafe and raises: []
• Remove the sink flag from the Outbox proc parameter

0.9.1

Bug Fixes

• Fix Lookup directives that contain a Not clause

0.9.0

Breaking Changes

No known breaking changes

New Features

• extend and join macros for combining tuple types
• eventSys system type
• Include import statements when using -d:dump
• Support for Optional components in Detach directives
• Add the Swap directive
• Add the -d:archetypes build flag
• Support component aliases with generics
• Save and Restore directives, along with saveSys and restoreSys system types
• Returning SystemInstance automatically flags a system as instanced.

Other Changes

• Ensure that archetype rows are never copied
• Silence noisy compiler warnings
• Reduce size of generated code by around 50%
• Logger is disabled by default

0.8.0

New Features

• Add the -d:necsusLog compiler flag for logging when systems are called
• Support for the -d:necsusFloat32 compiler flag to ensure float32 values are used internally
• Add a variant of SystemVar.getOrPut that sets default values
• Support sending events in from the outside world when an app instance is self-managed

Breaking Changes

• Removed the eventQueueSize parameter from newNecsusConf as it is now unused. Event queues are dynamically sized.
• Use a uint32 for tick IDs

Other Changes

• Support for ref types in system variables
• Removed code gen elision for nimsuggest as it was causing language server errors
• Remove unnecessary initializers in generated code
• Hosting of the readme directly on the doc site: https://necsusecs.github.io/Necsus/

0.7.0

Breaking Changes

1. Removed teardown and startup parameters from the necsus pragma. With this change, you should switch to using the startupSys and teardownSys pragmas attached directly to the systems instead. For example, this:

   import necsus
   proc startupSystem() = discard
   proc loopSystem() = discard
   proc teardownSys() = discard
   
   proc app() {.necsus([~startupSys], [~loopSys], [~teardownSys], newNecsusConf).}

   would become:

   import necsus
   proc startupSystem() {.startupSys.} = discard
   proc loopSystem() = discard
   proc teardownSys() {.teardownSys.} = discard
   
   proc app() {.necsus([~startupSys, ~loopSys, ~teardownSys], newNecsusConf).}

2. Spawn and Query no longer return an EntityId. If you need them, use a FullSpawn or a FullQuery instead. This change allows Necsus to improve build speeds and produce less output code. If you're interested in the details, read on.

   During a build, Necsus automatically generates a set of all possible archetypes that could possibly exist at runtime. It does this by examining systems with FullQuery, FullSpawn, Lookup, and Attach directives, then uses that to calculate all the combinatorial possibilities. Naively, this is an exponential algorithm. This is important because archetypes themselves aren't free. Each archetype that exists increases build times and slows down queries.

   Using Spawn instead of FullSpawn and Query instead of FullQuery allows the underlying algorithm to ignore those specific directives when calculating the final set of archetypes. Because your system doesn't have access to the EntityId, it can't use the output of a Spawn call as the input to an Attach directive, which means it can't contribute to the list of archetypes.

3. Convert TimeDelta and TimeElapsed from float to proc(): float. This means anywhere you were referencing it as a variable, you now need to invoke it as a proc. This was done to make sure any times stored in a Bundle return correct values.

New Features

1. Add the runSystemOnce macro. This makes it easier to test a system. This macro accepts a single lambda as an argument, and will invoke that lambda as if it were a system. You can then pass those directives to other systems, or interact with them directly.
2. Add a new directive, TickId. TickId gives you an auto-incrementing ID for each time a tick is executed. This is useful, for example, if you need to track whether an operation has already been performed for the current tick.
3. Add support for the -d:profile compiler flag. This allows you to get a quick and dirty idea of how your app is performing. When this flag is set, Necsus will add profiling code that reports how long each system is taking. It takes measurements, then outputs the timings to the console.

Other Changes

1. Allow Inboxes to work in a Bundle
2. Improve the code generated by -d:dump so it can be directly copied and pasted without change
3. Lazy archetype instantiation. This means that memory will only be allocated once an archetype is actually used
4. Use fewer closures when defining an app
5. Speed up archetype generation code by using a bitset instead of a full Set
6. Fix Bundles that use generics
7. Code gen an iterator for each query instead of using runtime generics
8. Various bug fixes
9. Various code-gen speed improvements

0.6.0

Breaking changes

Require archetype definitions to be sorted

This was done to simplify macro logic, which led to compile time performance improvvments.

Per system inboxes.

This fixes an unintuitive situation where an inbox appears in the system list before the outbox. In that case, events are sent, but never received because all the mailboxes were cleared at the end of every tick.

With this change, each system gets its own inbox, which is cleared after the system executes. It will mean more memory usage, but the difference should be small. And the behavior will be much more intuitive.

New Features

• Add state management via the active pragma. See readme for details
• Add SysVar.getOrPut
• Add Shared.isSome
• Add Inbox.len
• Bundle pragma. See readme for details

Other changes of note

• Many compile time speed improvements
• Various bug fixes
• Improved generic alias handling