An implementation of vector – dynamic linear array – in pure C89.
This one is competently generalized with macros (pseudo-templated), so you can create vector of any datatype supported in C – i.e. primitive types, structs and unions. Just preliminarily instantiate it for needed types and you're on.
Interface is based mostly on the design of std::vector
from C++11.
- Features in a nutshell
- Example (C99)
- Design notes
- Usage
- Functions
- Library adjustment using optional defines
- Access vector elements just like plain C arrays:
vec[k]
. - Support of multidimensional vectors (aka vector of vectors of...). Accessing them is Cpp-like too:
vec[i][j]
,vec[x][y][z]
, and so on. - It's possible to copy one vector into another, even if they contain values of different types.
- It's easy to instantiate necessary vector types once in a separate module, instead of doing this every time you needed a vector.
- You can choose how to pass values into a vector and how to return them from it: by value or by pointer.
- No code reduplication: only functions that take or return values of user type are specialized.
#include <stdlib.h>
#include <stdio.h>
#include "genvector/genvector.h"
typedef struct {
char Name[32];
int Age;
} person_s;
#define C_PERSON(name, age) ( (person_s){ .Name=name, .Age=age } )
GVEC_INSTANTIATE_EX( person_s, society, GENA_USE_SAMPLE, GENA_USE_ENTITY, GENA_ASSIGN_NAIVE );
int main() {
gvec_society_h family = gvec_society_new(0);
if (family == NULL) { return EXIT_FAILURE; }
gvec_society_push( &family, C_PERSON("Alice", 30) );
gvec_society_push( &family, C_PERSON("Bob", 32) );
gvec_society_push( &family, C_PERSON("Kate", 10) );
printf( "%zu\n", gvec_count( family ) );
while ( gvec_count( family ) > 0 ) {
person_s member = *gvec_society_pop( family );
printf( "name %s, age %d\n", member.Name, member.Age );
}
gvec_free( family );
return EXIT_SUCCESS;
}
- All indices and positions are zero-based.
- Functions validate their input arguments using only standard C89
assert()
from assert.h. - Non-specialized functions always return a defined result.
- If an error occurrs, the vector remains valid and unchanged.
- A vector storage never gets reduced, unless
gvec_shrink()
is called. - So-called "vector of void" type,
gvec_h
, is just a vector of untyped memory chunks.
By default, library provides only the next set of methods:
- Specialized functions.
- General-purpose functions.
- Instantiation macros.
To create a type "vector of T" and specialize management functions for it, you should instantiate it using instantiation macros. These are two and a half approaches in instantiation: static and modular, supplied with typesets. Let's examine them more closely.
This approach is good when vector is used only in one translation unit (module). It is easier and set by default.
Just include library header into module source and instantiate vector for types you need, using GVEC_INSTANTIATE()
:
GVEC_INSTANTIATE( tpTypeInfo, tpSurname, tpUseBy );
- tpTypeInfo – type for which vector should be instantiated
- tpSurname – unique vector name that will be placed into names of specialized functions, for example:
gvec_mystruct_new()
- tpUseBy – specifies how values should be passed to specialized functions and returned from them
Possible values both for tpUseBy are:
GENA_USE_SAMPLE
– pass/return by valueGENA_USE_ENTITY
– pass/return by reference
It is also a good practice to place library header inclusion and vector types instantiation in a separate header.
The main disadvantage about static approach is that vector type and it's corresponding specialized functions will be instantiated every time you use GVEC_INSTANTIATE()
. This is bad if same vector type is used in different modules – it will be instantiated for all of them, increasing output code size.
To prevent this problem, a modular approach should be used. Its idea is derived from the recommendation about separate header in the static approach: let's instantiate necessary vector types in a separate wrapper module, and use it instead of library itself every time you need a vector.
For the next code template let's assume that wrapper module is called gvec_wrapper.
gvec_wrapper.h
#include "genvector.h"
GVEC_H_DECLARE( tpTypeInfo, tpSurname, tpUseBy );
gvec_wrapper.c
#include "gvec_wrapper.h"
GVEC_C_DEFINE( tpTypeInfo, tpSurname, tpUseBy );
The arguments for GVEC_H_DECLARE()
and GVEC_C_DEFINE()
are the same as for GVEC_INSTANTIATE()
.
As you might have noticed, arguments for every pair of GVEC_H_DECLARE()
and GVEC_C_DEFINE()
are always the same. It's a sort of code duplication that may be considered undesirable.
To prevent this, typesets were introduced. Let's consider them using the modified version of the previous code template:
gvec_wrapper.h
#include "genvector.h"
#define ZZ_GVEC_TYPESET_SOMETHING \
(tpTypeInfo, tpSurname, tpUseBy)
GENA_APPLY_TYPESET( GVEC_H_DECLARE, ZZ_GVEC_TYPESET_SOMETHING );
gvec_wrapper.c
#include "gvec_wrapper.h"
GENA_APPLY_TYPESET( GVEC_C_DEFINE, ZZ_GVEC_TYPESET_SOMETHING );
Please note: gena_bool
type is fully compatible with bool
from stdbool.h in C99 and later, so it's preferred not to use gena_bool
if possible.
Notation:
NAME
: value oftpSurname
PASSVAL
:const tpTypeInfo
iftpUseBy
isGENA_USE_SAMPLE
, andconst tpTypeInfo*
otherwiseRETVAL
:tpTypeInfo
iftpUseBy
isGENA_USE_SAMPLE
, andtpTypeInfo*
otherwise
gvec_NAME_h gvec_NAME_new( size_t min_count )
Create a vector.
- min_count – a minimum count of elements that can be stored without storage relocation
Return value: a handle to the new vector, or NULL
on error
gena_bool gvec_NAME_reset( gvec_NAME_h* phandle, size_t count, const PASSVAL value )
Resize a vector to specified count of elements and assign a value to them all.
- phandle – a pointer to the handle to a vector
- count – a new count of elements in a vector
- value – a value to be assigned to elements
Return value: GENA_TRUE
if operation was performed successfully, GENA_FALSE
otherwise
gena_bool gvec_NAME_resize( gvec_NAME_h* phandle, size_t new_count, const PASSVAL value )
Resize a vector.
It's recommended to use gvec_resize()
for reducing the size, and gvec_NAME_resize()
for increasing.
- phandle – a pointer to the handle to a vector
- new_count – a new count of elements in a vector
- value – a value to be assigned to new elements
Return value: GENA_TRUE
if operation was performed successfully, GENA_FALSE
otherwise
gena_bool gvec_NAME_reserve( gvec_NAME_h* phandle, size_t min_count )
Reserve a space in a vector storage, at least for specified count of elements.
- phandle – a pointer to the handle to a vector
- min_count – a minimum count of elements that vector should accept without any relocation of its storage
Return value: GENA_TRUE
if operation was performed successfully, GENA_FALSE
otherwise
gena_bool gvec_shrink( gvec_NAME_h* phandle )
Free memory that isn't used by a vector now.
- phandle – a pointer to the handle to a vector
Return value: GENA_TRUE
if operation was performed successfully, GENA_FALSE
otherwise
gena_bool gvec_NAME_insert( gvec_NAME_h* phandle, size_t position, size_t count, const PASSVAL value )
Insert elements into a vector.
- phandle – a pointer to the handle to a vector
- position – a position of the first element to be inserted
- count – a count of elements to be inserted
- value – a value to be assigned to elements
Return value: GENA_TRUE
if operation was performed successfully, GENA_FALSE
otherwise
gena_bool gvec_NAME_push( gvec_NAME_h* phandle, const PASSVAL value )
Add an element to the end of a vector.
- phandle – a pointer to the handle to a vector
- value – a value to be assigned to the element
Return value: GENA_TRUE
if operation was performed successfully, GENA_FALSE
otherwise
RETVAL gvec_NAME_pop( gvec_h handle )
Drop the last element from a vector, and return it.
- handle – a handle to a vector
Return value:
tpUseBy
isGENA_USE_SAMPLE
: a value of the element (undefined if vector is empty)tpUseBy
isGENA_USE_ENTITY
: a pointer to the element, orNULL
if vector is empty
gvec_h gvec_assign( gvec_h* phandle, gvec_h source )
Copy-assign one vector to another. Sizes of the elements in both arrays must coincide. On error, the destination vector remains untouched.
- phandle – a pointer to the handle to a destination vector
- source – a handle to a source vector
Return value: a handle to the destination vector, or NULL
on error
gvec_h gvec_copy( gvec_h handle )
Duplicate a vector.
- handle – a handle to a source vector
Return value: a handle to the vector duplicate, or NULL
on error
void gvec_free( gvec_h handle )
Free a vector.
- handle – a handle to a vector (if
NULL
, nothing will occur)
void gvec_clear( gvec_h handle )
Clear all elements from a vector.
- handle – a handle to a vector
void gvec_reduce( gvec_h handle, size_t new_count )
Reduce a vector to the specified count of elements.
- handle – a handle to a vector
- new_count – a new count of elements in a vector (should not exceed the current count)
gena_bool gvec_remove( gvec_h handle, size_t position, size_t count )
Remove elements from a vector.
- handle – a handle to a vector
- position – a position of the first element to be removed
- count – a count of elements to be removed
Return value: GENA_TRUE
if operation was performed successfully, GENA_FALSE
otherwise
gena_bool gvec_drop( gvec_h handle )
Drop the last element from a vector.
- handle – a handle to a vector
Return value: GENA_TRUE
if operation was performed successfully, GENA_FALSE
otherwise
size_t gvec_count( gvec_h handle )
Get count of elements in a vector.
- handle – a handle to a vector
Return value: the count of elements
size_t gvec_size( gvec_h handle )
Get size of a vector storage.
- handle – a handle to a vector
Return value: the current size
gena_bool gvec_empty( gvec_h handle )
Returns if a vector specified is empty.
- handle – a handle to a vector
Return value: boolean
-
GVEC_GROWTH_FACTOR
(1.5 by default) Growth factor of vectors storages. -
GVEC_CALCULATE_SIZE_MATH
Use math function for size calculation, instead of loop-based. -
GVEC_INSERT_NO_REALLOC
Don't perform storage relocation usingrealloc()
on elements insertion. This prevents excess memory copying when inserting elements not at the end of a vector.
It's also recommended to compile with NDEBUG
defined, to disable assertion checks.