Differences

This shows you the differences between two versions of the page.

--- en:toolworks:docs:apparatus:operating [2021/08/28 23:59] – created vladius
+++ en:toolworks:docs:apparatus:operating [2023/01/14 13:13] (current) – vladius
@@ Line 1: / Line 1: @@
 ====== Operating ======
-The newer and more robust way of processing your chains is through the process called operating.
+The newer and more robust way of handle your chains is through the process called operating.
 ===== C++ Workflow =====
@@ Line 7: / Line 7: @@
 ==== Using a Lambda ====
-You can easily operate on your chain via a C++ lambda and this is how you do it:
+You can easily operate on your chain via a [[cpp>language/lambda|C++ lambda]] and this is how you do it:
 <code cpp>
-Chain->Operate([](const FChain::FCursor& Cursor, FMyTrait Trait)
+Chain->Operate([](FMyTrait Trait)
 {
     ...
@@ Line 15: / Line 15: @@
 </code>
-The type of cursor here must match the type of the chain used.
 Note that you're not allowed to acquire a reference to the trait while processing a non-solid chain, only its copy.
 So in order to operate on a solid chain, you could do something like this:
 <code cpp>
-SolidChain->Operate([](const FSolidChain::FCursor& Cursor, FMyTrait& Trait)
+SolidChain->Operate([](FMyTrait& Trait)
 {
     ...
@@ Line 26: / Line 25: @@
 </code>
-Now you can change the properties (fields) of the trait directly, without copying invloved.
+Now you can change the properties (fields) of the trait directly, without copying involved.
+==== Concurrency ====
+Solid Chains also support a special type of operating - a multi-threaded one.
+The function to call is explicitly named with a ''Concurrently'' prefix and
+accepts two more arguments: the maximum number of tasks to utilize and the minimum number
+of slots per each such task. For example:
+<code cpp>
+SolidChain->OperateConcurrently([](FMyTrait& Trait)
+{
+    ...
+}, 4, 32);
+</code>
+The second parameter helps to also limit the number of tasks. If there are too little slots
+available, excessive tasks not needed for that quantity won't be queued at all.
+==== Dependency Injection ====
+One great thing about operating is that the function arguments are actually resolved
+and delivered automatically to your logic. For example, if you also modify the currently
+iterated subject, just specify the Subject handle in the very declaration of the routine:<code cpp>
+Chain->Operate([](FSubjectHandle Subject, FMyTrait Trait)
+{
+    ...
+});
+</code>
+This, of course, has to match the solidity of the chain. So for a solid chain this would be:<code cpp>
+SolidChain->Operate([](FSolidSubjectHandle Subject, FMyTrait& Trait)
+{
+    ...
+});
+</code>
+You can actually ask for different contextual information within the loop. For example:<code cpp>
+Chain->Operate([](const FChain* Chain, const FChainCursor& Cursor, ISubjective* Subjective, FMyTrait Trait, UMyDetail* Detail)
+{
+    ...
+});
+</code>
+==== Current Iteration Index ====
+Just in case you need to know the number of the current iterated Slot (i.e. Subject's place within the chain), use the dedicated [[appi>struct_t_chain_1_1_f_cursor.html#a18005eddb3a624bdabb76475c3863751|GetChainSlotIndex()]] method of the corresponding Cursor type, which can also be delivered using the [[#argument_delivery|aforementioned means]]:<code cpp>
+SolidChain->Operate([](FPlacementTrait& Placement, const FSolidChainCursor& Cursor)
+{
+	Placement.Number = Cursor.GetChainSlotIndex();
+});
+</code>
+==== Stopping ====
+While conceptually not very clean it is sometime useful to stop the actual processing (iterating) of the chain
+prematurely, manually. This can be easily accomplished with a dedicated [[appi>struct_t_chain.html#aec97fe73be3d6d9f4c279c7427ed99d6|method]].
+For example:<code cpp>
+int32 Counter = 0;
+Chain->Operate([&Counter](const FChain* Chain, FMyTrait Trait)
+{
+    if (Counter > 100)
+    {
+        Chain->StopIterating();
+        // Return explicitly, so the counter doesn't get incremented on the current iteration:
+        return;
+    }
+    Counter += Trait.Value;
+});
+</code>
+==== Direct Mechanism Operating ====
+The Operating simplification goes as far as actually Operating on a Mechanism directly, like so:<code cpp>
+Mechanism->Operate([](FGlowing Glowing, FHelmet Helmet)
+{
+    ...
+});
+</code>
+This way the [[en:toolworks:docs:apparatus:filter|Filtering]] and the [[en:toolworks:docs:apparatus:enchaining|Enchaining]] is done automatically under the hood, deriving the necessary Components from the lambda arguments.
+You can of course supply the Filter specification explicitly, overriding it.
+So for example, if you want to specify an additional [[en:toolworks:docs:apparatus:trait|Trait]] and a [[en:toolworks:docs:apparatus:flag|Flagmark]] condition, do it like so:<code cpp>
+Mechanism->Operate(FFilter::Make<FGlowing, FHelmet, FHero>(FM_Z), [](FGlowing Glowing, FHelmet Helmet)
+{
+    ...
+});
+</code>
+The direct Operating mode is smart enough to deduce the type of the Chain used within the Operating process, so if you specify a reference to a Trait and/or a Solid Subject handle in your arguments list, the technique will essentially produce a [[en:toolworks:docs:apparatus:solidity|Solid]] iterating:<code cpp>
+Mechanism->Operate([](FSolidSubjectHandle Subject, FGlowing& Glowing, FHelmet& Helmet)
+{
+    ...
+});
+</code>
+You can still specify the Chain type explicitly as a first template argument to a method:<code cpp>
+Mechanism->Operate<FSolidChain>([](FGlowing Glowing, FHelmet Helmet)
+{
+    // You logic within Solid semantics:
+    ...
+});
+</code>
+Concurrency variants are also provided by the direct interface:<code cpp>
+Mechanism->OperateConcurrently([](FSolidSubjectHandle Subject, FGlowing& Glowing, FHelmet& Helmet)
+{
+    ...
+}, /*Maximum number of threads=*/4, /*Minimum number of Subjects=*/16);
+</code>