I want to follow up on this question: What is the difference between a definition and a declaration?, there is answer about definition and a declaration, but nothing about implementation.
I want also know what is the difference between implementation and definition.
implementation: if you have some pseudo-code or something like an UML-Diagram and write your code on that basis it's an implementation
declaration: a declaration is already in your code where you say the compiler/interpreter: "hey look there is this variable that I want to use but I dont want to give it any value yet"
and finally definition: definition is when you finally assign a value to your variable like x = 4. Like your defining x to be 4 (in your code)
Hope this is helpful to you
Function definition etc is a formal term, but "implementation" is a fuzzy informal term. In plain English, it could refer to your application's implementation of something, like for example a function. The implementation phase of a project is typically the phase where you write all the code. And so on - it depends on context.
In formal/technical C programming terms, implementation means the implementation of the C language. That is: formally implementation means the C compiler and standard library. And the formal term implementation-defined behavior means compiler-specific behavior.
Related
I know that C is not a context-free language, a famous example is:
int foo;
typedef int foo;
foo x;
In this case the lexer doesn't know, whether foo in the 3rd line, is an identifier, or typedef.
My question is, is this the only reason that makes C a Context-Sensitive Language?
I mean, if we get rid of typedef, would it become context-free language? Or there are other reasons (examples) that prevent it from being so?
Yes. C can be parsed with a classical lex + yacc combo. The lexer definition and the yacc grammar are freely available at
http://www.quut.com/c/ANSI-C-grammar-l-2011.html
and
http://www.quut.com/c/ANSI-C-grammar-y-2011.html
As you can see from the lex file, it's straightforward except for the context-sensitive check_type() (and comment(), but comment processing technically belongs to the preprocessor), which makes typedef the only source of context-sensitivity there. Since the yacc file doesn't contain any context-sensitivity introducing tricks either, a typedef-less C would be a perfectly context-free language.
No. C cannot be a strict context independent language. For that, you should describe a syntax that doesn't allow to use a nondeclared variable (this is context) in a similar way as what you describe in your question. The language authors always describe syntax using some kind of context free grammar, but just to describe the main syntactic constructs of the language. The case you describe (making a type identifier to fit in a different token class to be able to go in places where it shouldn't) is only an example. If you look for example, the freedom in the order for things like static unsigned long long int variable simplifies the syntax remembering by programmers, but complicates things to the compiler authors.
As per my knowledge and research there are two basic reasons that make C context sensitive language. These are:
Variable is declared before it is used.
Matching the formal and actual parameters of functions or sub-routines.
These two can't be done by PushDown Automata (PDA) but Linear Bounded Automata (LBA) can do thes two.
I have code that gives me an error. Implicit declaration of isNumericFloat.
I want to know if the function:
isNumericFloat()
a built it function in C?
NO, it's not a "built-in" c function.1
This function is used somewhere in your code and it's not part of the standard library. In fact, just because it uses camel case which is not very common in c code it seems like an odd function written by a not so c-ish programmer, of course that's a subjective reason, but commonly c programmers would choose is_numeric_float().
You need to search your code to see if you can find it's defintion, but in the mean time you can provide a prototype, like
int isNumericFloat(float value); // I don't really know what arguments it takes
// but you can surely infer them from the code
before it's ever called in the code, if you do so one of these two things will happen
If there is a definition for the function somewhere, it will compile fine.
If there is no definition, the linker will tell you that there is/are undefined reference/s to it in the code.
1Strictly speaking, there are no built-in functions in c, there is something called the standard library (headers starting with std , like stdlib.h), and I mean that it's not part of such library.
I know that C uses pass-by-value and we can emulate pass-by-reference with the help of pointers. But, for example, in order to calculate a simple mathematical expression, how do I implement pass-by-name (which is kind of lazy evaluation but not exactly) in C?
C is only pass-by-value. You can't pass by reference or name. With the pre-processor you can do various hacks but not in the C language.
Sometimes, people call passing a pointer "pass-by-reference" but this is not the case. The pointer is passed by value like anything else. C++ is a different story but you asked about C.
You might also be interested in this article discussion this at length
The parameter substitution used by function-like preprocessor macros is sometimes described as being "pass by name".
Is it legal to forward declare structs and functions provided by the C standard library?
My background is C++ in which the answer is no. The primary reason for this is that a struct or class mandated by the C++ standard library can be a template behind the scenes and may have "secret" template parameters and so cannot be properly declared with a naive non-template declaration. Even if a user does figure out exactly how to forward declare a particular entity in a particular version of a particular implementation, the implementation is not obliged to not break that declaration in future versions.
I don't have a copy of any C standard at hand but obviously there are no templates in C.
So is it legal to forward declare entities in the C standard library?
Another reason that entities in the C++ standard library may not be forward declared is that headers provided by the implementation need not follow the normal rules. For example, in a recent question I asked if a C++ header provided by the implementation need be an actual file and the answer was no. I don't know if any of that applies to C.
The C standard library is used by both C and C++ but for this question I'm only asking about C.
Forward declarations of structs are always permissible in C. However, not very many types can be used this way. For example, you can't use a forward declaration for FILE simply because the tag name of the struct is not specified (and theoretically, it may not be a struct at all).
Section 7.1.4 paragraph 2 of n1570 gives you permission to do the same with functions:
Provided that a library function can be declared without reference to any type defined in a
header, it is also permissible to declare the function and use it without including its
associated header.
This used to be rather common. I think the reasoning here is that hard drives are slow, and fewer #include means faster compile times. But this isn't the 1980s any more, and we all have fast CPUs and fast hard drives, so a few #include aren't even noticed.
void *malloc(size_t);
void abort(void);
/* my code here */
yes you can this is perfectly valid.
this can be done with the standard library too.
double atof(const char *);
int main() {
double t = atof("13.37");
return 0;
}
#include <stdio.h>
Similiar things can be done with structs, variables etc.
I would recommend you read the wiki page which features some c examples:
http://en.wikipedia.org/wiki/Forward_declaration
this is specified in the c standard, Section 7.1.4 paragraph 2 of n1570
Provided that a library function can be declared without reference to any type defined in a header, it is also permissible to declare the function and use it without including its associated header.
Let's say I have a macro in an inclusion file:
// a.h
#define VALUE SUBSTITUTE
And another file that includes it:
// b.h
#define SUBSTITUTE 3
#include "a.h"
Is it the case that VALUE is now defined to SUBSTITUTE and will be macro expanded in two passes to 3, or is it the case that VALUE has been set to the macro expanded value of SUBSTITUTE (i.e. 3)?
I ask this question in the interest of trying to understand the Boost preprocessor library and how its BOOST_PP_SLOT defines work (edit: and I mean the underlying workings). Therefore, while I am asking the above question, I'd also be interested if anyone could explain that.
(and I guess I'd also like to know where the heck to find the 'painted blue' rules are written...)
VALUE is defined as SUBSTITUTE. The definition of VALUE is not aware at any point that SUBSTITUTE has also been defined. After VALUE is replaced, whatever it was replaced by will be scanned again, and potentially more replacements applied then. All defines exist in their own conceptual space, completely unaware of each other; they only interact with one another at the site of expansion in the main program text (defines are directives, and thus not part of the program proper).
The rules for the preprocessor are specified alongside the rules for C proper in the language standard. The standard documents themselves cost money, but you can usually download the "final draft" for free; the latest (C11) can be found here: http://www.open-std.org/jtc1/sc22/wg14/www/docs/n1570.pdf
For at-home use the draft is pretty much equivalent to the real thing. Most people who quote the standard are actually looking at copies of the draft. (Certainly it's closer to the actual standard than any real-world C compiler is...)
There's a more accessible description of the macro rules in the GCC manual: http://gcc.gnu.org/onlinedocs/cpp/Self_002dReferential-Macros.html
Additionally... I couldn't tell you much about the Boost preprocessor library, not having used it, but there's a beautiful pair of libraries by the same authors called Order and Chaos that are very "clean" (as macro code goes) and easy to understand. They're more academic in tone and intended to be pure rather than portable; which might make them easier reading.
(Since I don't know Boost PP I don't know how relevant this is to your question but) there's also a good introductory example of the kids of techniques these libraries use for advanced metaprogramming constructs in this answer: Is the C99 preprocessor Turing complete?