DIP55

From D Wiki
Revision as of 20:30, 3 January 2014 by FBergemann (talk | contribs) (Compositions via the stack)
Jump to: navigation, search
Title: Access Data Items In Ancestor Stack Frames
DIP: 55
Version: 1
Status: Draft (under construction)
Created: 2014-01-01
Last Modified: 2014-01-01
Author: Frank Bergemann
Links: Dr.Dobb's Article Access Data Items In Ancestor Stack Frames Safely

NG discussion for C++

Abstract

The language should support an implicit *caller pointer for functions (callee) for access to the calling function (caller).

As if functions would internally be modelled as class wrappers (functors) passing their *this ptr implicitly to a called function.

(Likely this behavior is switched on/off, auto-detected by making use of the *caller ptr in a called function yes/no.)

Rationale

callee pulls data specificially when/if required instead of caller pushs all always

This might be a performance improvement.

f1(void)
{
   SomeType x;
   SomeType y;
   SomeType z;
   [...]
   f2();
   [...]
}

f2(void)
{
   if (caller->x.IsValid())
   {
      use(caller->y)
   }
   else 
   {
      use(caller->z);
   }
}

public, private (protected?) caller data

Like for data members for classes, local variables of functions might be declared private, public (protected?) for access limitations

f1(void)
{
private:
   int _x;
   int _y;
public:
   int z;
   [...]
   f2();
   [...]
}

f2(void)
{
private: 
   int _a1 = caller->_x * 2.0; // ERROR!
   int _a2 = caller->z * 2.0;  // OK
   [...]
}

non-const for the caller, const for the callee

f1(void)
{
private:
   int _x;
   int _y;
public:
   const int & x = _x;
   const int & y = _y;
   [...]
   f2();
   [...]
}

f2(void)
{
private: 
   caller->x *=  2.0;        // ERROR!
   int x2 = caller->x * 2.0; // OK
   [...]
}

public, private (protected?) local functions

Such might serve as getter or setters with similar benefits we have for accessing class data.

f1(void)
{
private:
   SomeProtoType _x;
public:
   SomeType GetX()(void)
   {
      return SomeType(_x);
   }

   [...]
   f2();
   [...]
}

f2(void)
{
   if (whatsoever())
   {
      DoIt(caller->GetX());
   }
}

E.g....

Lazy evaluation

Data could be prepared only by caller for callee. But it could be up to the callee to decide if it actually wants that data and for this delay its calculation.

f1(void)
{
   SomeType x;
   SomeType y;
   SomeType z;
   [...]
   f2();
   [...]
}

f2(void)
{
   if (caller->x.IsValid())
   {
      caller->y.ExpensivePreProcess();
      use(caller->y);
   }
   else 
   {
      caller->z.ExpensivePreProcess();
      use(caller->z);
   }
}

Multi-level

f1(void)
{
public:
   RootData rootData;
private:
   [...]
   f2();
   [...]
}

f2(void)
{
public:
   SomeData x;
   SomeData y;
   SomeData z;

   f3();
}

f3(void)
{
   if (caller->x.IsValid())
   {
      use(caller->caller->rootData, caller->y)
   }
   else 
   {
      use(caller->caller->rootData, caller->z)
   }
}

plus a *callscope ptr?

caller->caller->x has a flaw: We need to know about the call hierarchy (levels - is it #1 level up or #2 levels up?).

Can't we have something like callscope->x which check for the closest x in ancestor stack frames?

This would be compatible with existing definitions for shadowing:

funcA(...)
{
   int x = 1;

   funcB(x); // x is #1
}

funcB(...)
{
   // funcB body
}


funcA(...)
{
   int x = 1;

   {
      int x = 2;
      funcB(x); // x is #2
   }
}

funcB(...)
{
   // funcB body
}

unrolled:

funcA(...)
{
   int x = 1;

   {
      int x = 2;
      {
         // funcB body // x is #2
      }
   }
}

The latter (unrolled) reveals that this CR moves functions a bit closer to local scopes. Because the access to "surrounding data" is similar.

Cons

caller has to adhere to callee's naming

For passing data as function arguments, the function can choose any name w/o dependency to the caller. The caller just provides the values (but has to follow a sequence).

However, when the called function tries to use some caller->x, it expects the caller to provide some 'x'.

On the other side: if using named parameters, the caller also has be aware of the callee's naming conventions.

But for this the caller still can decide for the function call invocation which value to pass as 'x => someVal'. But does not have to set a variable (pointer, reference), which does this.

Description

Dependency and Hierarchy

Function A is calling function B.

Usually A has more domain knowledge. And B is a less informed service for A.

But can't B have more limited, but more precise information about a certain issue?

Could that be a reason, to turn B from a stupid service into an instructor for A?

Is it good or a bad design to let B directly manipulate A?

E.g.:

1. terminate B1 (itself), but install an (alternate) B2 for processing next?

2. replace B1 (itself) by B2A and B2B (split)

Or should B always only return low level information to A and leave it up to A to decide what to do next?

Usage

STILL TEMPLATE CONTENT ONLY HERE

This section has been adapted for MediaWiki.

To start a new DIP you can go to Edit link and copy the source of this DIP, then go to DIP index and add a new item in the list. The DIP number should be one more than the last DIP in the index (for example, if the DIP1 is the last DIP, your DIP should be DIP2). The link in the index should have the form: [[DIPx]], Title, Status, resume. Where x is the DIP number, title is the DIP title and resume is a short description about the DIP.

Save the DIP index page and click on the new red link. Now you are editing the new DIP you just created, now paste the copied source text from this template and replace all the you need.

Remember to update the metadata at the start of the DIP, and keep it as a Draft at the beginning. When your DIP is done, you should announce it in the News Group for discussion, with a subject like this: new DIPx: title (where one more time x is the DIP number and title is the DIP title).

You should always put you DIPs in the Public Domain (or a similarly permissive license but use Public Domain unless you're very sure of what you're doing).

Recommendations

STILL TEMPLATE CONTENT ONLY HERE

When writing a DIP, try not to express your opinion. DIPs should provide facts and be as objective as possible. Even when that's pretty hard, you can make the DIP look more objective by not using, for example, "I prefer XXX because YYY". If YYY is an objective advantage, write that in a more objective way, like "XXX can be a better option because YYY". Try to leave non-technical personal preferences aside; "XXX can be a better option because the syntax is nicer" is not good enough even when you don't say "I prefer".

Try not to include half-baked ideas. If you are not sure about something, leave it outside the DIP and write it on the NG announcement instead for further discussion. The idea can be added to the DIP in the future when it is in a better shape.

Abstract

What's good for classes is good for functions

This CR enables features used for classes (structural) to functions/procedures (procedural).

Bypassing data

As it shares a context (data) along a call hierarchy it can be considered a "better version of global variables" - because it is limited to the execution scope (call hierarchy).

How many times you've been bothered to pass some value along a call hierarchy to reach a target function?

Now you can tunnel it.

FuncA(...)
{
   SomeType transactionId;

   FuncB(...); // NOT including transactionId
}

FuncB(...)
{
   // not using FuncA's transactionId;
   FuncC(...); 
}

FuncC(...)
{
   Log(callscope>transactionId, "my message");
}

But don't stress this option - don't end up in a BLOB stack.

For the example above: you need some logging function - even don't want to know requirement to use transactionId for logging?

FuncA(...)
{
   SomeType transactionId;

   void Log(...)
   {
     writeln(transactionId):
     foreach (arg; _arguments)
     {
	[...]
     }
   }

   FuncB(...); // NOT including transactionId
}

FuncB(...)
{
   // not using FuncA's transactionId;
   FuncC(...); 
}

FuncC(...)
{
   callscope->Log("my message");
}

Closure

FuncA(void)
{
   NestedFuncB(void)
   {
   public: 
      // selected context to flavour free function DoRealWork()
      int x = 1;
      int y = 2;
      DoRealWork();
   }

   [...]
}

DoRealWork(void)
{
   // do whatever you like using...
   caller->x
   caller->y
}

Compositions via the stack

Inheritance is used to model basic features in base classes and to specialize or to add extension in derived classes.

This is done by design before/at compile time.

Turning functions into some class-alike for this CR provides inheritance-alike options along the stack.

An initial function FuncA(...) can hold a base definition (data, nested functions). And invoking a next FuncB(...) can add extensions.

At FuncX(...) some desired context is reached.

Note: such shall not replace class design, it's just a different tool.

In contrast to a structural class using a stack of functions is not durable, but rather temporary (as a start). But usable for a singleton-alike work-horse (processor).

But in contrast to classes, such context can be composed dynamically (at runtime). And it can be composed limited to the current needs. E.g. a base functionality can be completed with an import interface at one time, and later the import interface can be dropped and an export interface can be added.

Video Inheritance is the base class of Evil might address the same issue (i am not sure).

Intentionally deal with the stack

It offers a toolbox to make better use of a stack. It adds a stack-machine user interface for the programmer (which might be usable for compilers?).

E.g. for traversing trees the stack can be better incorporated into the algorithm implementation. A parent/child relationship is supported by a caller/called model.

(see e.g. ASF RBT as a trial to make use of this programming model for a Red/Black-Tree algorithm - w/o need for a parent ptr in the node structure (however it is incomplete - the delete operation is still missing))

Rationale

STILL TEMPLATE CONTENT ONLY HERE

Rationale should be complete. When the DIP tries to solve a problem, try to describe that problem as detailed as possible. If you have links to the NG describing the problem more deeply, used them. All the background information is welcome.

NG Announcement

STILL TEMPLATE CONTENT ONLY HERE

When posting the DIP announcement to the NG, please copy the abstract, so people can easily know what is it about and follow the link if they are interested.

Copyright

This document has been placed in the Public Domain.