Bored with developing the same functionality over and over again, unsatisfied with existing libraries, so decided to make the right one, once and forever. 🙂
- Handles both C and binary strings;
- Light-weight references to strings: cheap to create, copy, or pass by value;
- Support for copy and move semantics, although not enforceable by the C language;
- String composition functions writing to memory, file descriptors, or file streams;
- Can be compiled using
gcc
orclang
, and linked withlibc
ormusl
.
Just clone the project and copy (or symlink) the files str.h
and str.c
into your project,
but please respect the license.
String composition:
str s = str_null;
str_join(&s, str_lit(", "),
str_lit("Here"),
str_lit("there"),
str_lit("and everywhere"));
str_cat(&s, s, str_lit("..."));
assert(str_eq(s, str_lit("Here, there, and everywhere...")));
str_free(s);
Same as above, but writing to a file:
FILE* const stream = fopen(...);
int err = str_join(stream, str_lit(", "),
str_lit("Here"),
str_lit("there"),
str_lit("and everywhere..."));
if(err != 0) { /* handle the error */ }
Discussion on Hacker News.
Disclaimer: This is the good old C language, not C++ or Rust, so nothing can be enforced on the language level, and certain discipline is required to make sure there is no corrupt or leaked memory resulting from using this library.
A string is represented by the type str
that maintains a pointer to some memory containing the
actual string, and the length of the string. Objects of type str
are small enough (a struct
of a const char*
and a size_t
) to be cheap to create, copy (pass by value), and move. The
str
structure should be treated as opaque (i.e., do not attempt to directly access or modify
the fields in this structure). The strings are assumed to be immutable, like those in Java or
Go, but only by means of const char*
pointers, so it is actually possible to modify such a
string, although the required type cast to char*
offers at least some (mostly psychological)
protection from changing the string by mistake.
This library focusses only on handling strings, not gradually composing them like StringBuffer class in Java.
All string objects must be initialised before use. Uninitialised objects will cause
undefined behaviour. Use the provided constructors, or str_null
for empty strings.
There are two kinds of str
objects: those actually owning the memory they point to, and
non-owning references. This property can be queried using str_is_owner
and str_is_ref
functions, otherwise such objects are indistinguishable.
Non-owning string objects are safe to copy and assign to each other, as long as the memory
they refer to is valid. They do not need to be freed. str_free
is a no-op for reference
objects. A reference object can be cheaply created from a C string, a string literal,
or from a range of bytes.
Owning objects require special treatment, in particular:
- It is a good idea to have only one owning object per each allocated string, but such
a string can have many references to its underlying string, as long as those references do not
outlive the owning object.
Sometimes this rule may be relaxed for code clarity, like in the above example where
the owning object is passed directly to a function, but only if the function does not
store or release the object. When in doubt pass such an object via
str_ref
. - Direct assignments (like
s2 = s1;
) to owning objects will certainly leak memory, usestr_assign
function instead. In fact, this function can assign to any string object, owning or not, so it can be used everywhere, just to avoid any doubt. - There is no automatic memory management in C, so every owning object must be released at
some point using either
str_free
orstr_clear
function. String objects on the stack can also be declared asstr_auto
(orconst str_auto
) for automatic cleanup when the variable goes out of scope. - An owning object can be moved to another location by using
str_move
function. The function resets its source object to an empty string. - Object ownership can be passed over to another object by using
str_pass
function. The function sets its source to a non-owning reference to the original string.
It is technically possible to create a reference to a string that is not null-terminated. The library accepts strings without null-terminators, but every new string allocated by the library is guaranteed to be null-terminated.
A string object can be constructed form any C string, string literal, or a range of bytes.
The provided constructors are computationally cheap to apply. Depending on the constructor,
the new object can either own the actual string it refers to, or be a non-owning reference.
Constructors themselves do not allocate any memory. Importantly, constructors are the only
functions in this library that return a string object, while others only assign their results
through a pointer to a pre-existing string. This makes constructors suitable for initialisation
of new string objects. In all other situations one should combine construction with assignment,
for example:
str_assign(&dest, str_acquire_chars(buff, n));
Querying a property of a string object (like the length of the string via str_len
) is a
cheap operation.
C language does not allow for operator overloading, so this library provides a function
str_assign
that takes a string object and assigns it to the destination object, freeing
any memory owned by the destination. It is generally recommended to use this function
everywhere outside object initialisation.
An existing object can be moved over to another location via str_move
function.
The function resets the source object to str_null
to guarantee the correct move semantics.
The value returned by str_move
may be either used to initialise a new object, or
assigned to an existing object using str_assign
.
An existing object can also be passed over to another location via str_pass
function. The function
sets the source object to be a non-owning reference to the original string, otherwise the semantics
and usage is the same as str_move
.
String composition functions can write their results to different
destinations, depending on the type of their dest
parameter:
str*
: result is assigned to the string object;int
: result is written to the file descriptor;FILE*
result is written to the file stream.
The composition functions return 0 on success, or the value of errno
as retrieved at the point
of failure (including ENOMEM
on memory allocation error).
Just to make things more clear, here is the same code as in the example above, but with comments:
// declare a variable and initialise it with an empty string; could also be declared as "str_auto"
// to avoid explicit call to str_free() below.
str s = str_null;
// join the given string literals around the separator (second parameter),
// storing the result in object "s" (first parameter); in this example we do not check
// the return values of the composition functions, thus ignoring memory allocation failures,
// which is probably not the best idea in general.
str_join(&s, str_lit(", "),
str_lit("Here"),
str_lit("there"),
str_lit("and everywhere"));
// create a new string concatenating "s" and a literal; the function only modifies its
// destination object "s" after the result is computed, also freeing the destination
// before the assignment, so it is safe to use "s" as both a parameter and a destination.
// note: we pass a copy of the owning object "s" as the second parameter, and here it is
// safe to do so because this particular function does not modify its arguments.
str_cat(&s, s, str_lit("..."));
// check that we have got the expected result
assert(str_eq(s, str_lit("Here, there, and everywhere...")));
// finally, free the memory allocated for the string
str_free(s);
There are some useful code snippets provided to assist with writing code using this library.
typedef struct { ... } str;
The string object.
size_t str_len(const str s)
Returns the number of bytes in the string referenced by the object.
const char* str_ptr(const str s)
Returns a pointer to the first byte of the string referenced by the object. The pointer is never NULL.
const char* str_end(const str s)
Returns a pointer to the next byte past the end of the string referenced by the object.
The pointer is never NULL, but it is not guaranteed to point to any valid byte or location.
For C strings it points to the terminating null character. For any given string s
the following
condition is always satisfied: str_end(s) == str_ptr(s) + str_len(s)
.
bool str_is_empty(const str s)
Returns "true" for empty strings.
bool str_is_owner(const str s)
Returns "true" if the string object is the owner of the memory it references.
bool str_is_ref(const str s)
Returns "true" if the string object does not own the memory it references.
str_null
Empty string constant.
str str_lit(s)
Constructs a non-owning object from a string literal. Implemented as a macro.
str str_ref(s)
Constructs a non-owning object from either a null-terminated C string, or another str
object.
Implemented as a macro.
str str_ref_chars(const char* const s, const size_t n)
Constructs a non-owning object referencing the given range of bytes.
str str_acquire_chars(const char* const s, const size_t n)
Constructs an owning object for the specified range of bytes. The pointer s
should be safe
to pass to free(3)
function.
str str_acquire(const char* const s)
Constructs an owning object from the given C string. The string should be safe to pass to
free(3)
function.
str str_move(str* const ps)
Saves the given object to a temporary, resets the source object to str_null
, and then
returns the saved object.
str str_pass(str* const ps)
Saves the given object to a temporary, sets the source object to be a non-owning reference to the
original string, and then returns the saved object.
void str_free(const str s)
Deallocates any memory held by the owning string object. No-op for references. After a call to
this function the string object is in unknown and unusable state.
String objects on the stack can also be declared as str_auto
instead of str
to deallocate
any memory held by the string when the variable goes out of scope.
void str_assign(str* const ps, const str s)
Assigns the object s
to the object pointed to by ps
. Any memory owned by the target
object is freed before the assignment.
void str_clear(str* const ps)
Sets the target object to str_null
after freeing any memory owned by the target.
void str_swap(str* const s1, str* const s2)
Swaps two string objects.
int str_from_file(str* const dest, const char* const file_name)
Reads the entire file (of up to 64MB by default, configurable via STR_MAX_FILE_SIZE
) into
the destination string. Returns 0 on success, or the value of errno
on error.
int str_cmp(const str s1, const str s2)
Lexicographically compares the two string objects, with usual semantics.
bool str_eq(const str s1, const str s2)
Returns "true" if the two strings match exactly.
int str_cmp_ci(const str s1, const str s2)
Case-insensitive comparison of two strings, implemented using strncasecmp(3)
.
bool str_eq_ci(const str s1, const str s2
Returns "true" is the two strings match case-insensitively.
bool str_has_prefix(const str s, const str prefix)
Tests if the given string s
starts with the specified prefix.
bool str_has_suffix(const str s, const str suffix)
Tests if the given string s
ends with the specified suffix.
int str_cpy(dest, const str src)
Copies the source string referenced by src
to the
generic destination dest
. Returns 0 on success,
or the value of errno
on failure.
int str_cat_range(dest, const str* src, size_t count)
Concatenates count
strings from the array starting at address src
, and writes
the result to the generic destination dest
.
Returns 0 on success, or the value of errno
on failure.
int str_cat(dest, ...)
Concatenates a variable list of str
arguments, and writes the result to the
generic destination dest
.
Returns 0 on success, or the value of errno
on failure.
int str_join_range(dest, const str sep, const str* src, size_t count)
Joins around sep
the count
strings from the array starting at address src
, and writes
the result to the generic destination dest
.
Returns 0 on success, or the value of errno
on failure.
int str_join(dest, const str sep, ...)
Joins a variable list of str
arguments around sep
delimiter, and writes the result to the
generic destination dest
.
Returns 0 on success, or the value of errno
on failure.
bool str_partition(const str src, const str patt, str* const prefix, str* const suffix)
Splits the string src
on the first match of patt
, assigning a reference to the part
of the string before the match to the prefix
object, and the part after the match to the
suffix
object. Returns true
if a match has been found, or false
otherwise, also
setting prefix
to reference the entire src
string, and clearing the suffix
object.
Empty pattern patt
never matches.
void str_sort_range(const str_cmp_func cmp, str* const array, const size_t count)
Sorts the given array of str
objects using the given comparison function. A number
of typically used comparison functions is also provided:
str_order_asc
(ascending sort)str_order_desc
(descending sort)str_order_asc_ci
(ascending case-insensitive sort)str_order_desc_ci
(descending case-insensitive sort)
const str* str_search_range(const str key, const str* const array, const size_t count)
Binary search for the given key. The input array must be sorted using str_order_asc
.
Returns a pointer to the string matching the key, or NULL.
size_t str_partition_range(bool (*pred)(const str), str* const array, const size_t count)
Reorders the string objects in the given range in such a way that all elements for which
the predicate pred
returns "true" precede the elements for which predicate pred
returns "false". Returns the number of preceding objects.
size_t str_unique_range(str* const array, const size_t count)
Reorders the string objects in the given range in such a way that there are two partitions:
one where each object is unique within the input range, and another partition with all the
remaining objects. The unique partition is stored at the beginning of the array, and is
sorted in ascending order, followed by the partition with all remaining objects.
Returns the number of unique objects.
for_each_codepoint(var_name, src_string)
A macro that expands to a loop iterating over the given string src_string
(of type str
) by UTF-32
code points. On each iteration the variable var_name
(of type char32_t
) is assigned
the value of the next valid UTF-32 code point from the source string. Upon exit from the loop the
variable has one on the following values:
CPI_END_OF_STRING
: the iteration has reached the end of source string;CPI_ERR_INCOMPLETE_SEQ
: an incomplete byte sequence has been detected;CPI_ERR_INVALID_ENCODING
: an invalid byte sequence has been detected.
The source string is expected to be encoded in the current program locale, as set by the most
recent call to setlocale(3)
.
Usage pattern:
#include <uchar.h>
...
str s = ...
...
char32_t c; // variable to receive UTF-32 values on each iteration
for_each_codepoint(c, s)
{
/* process c */
}
if(c != CPI_END_OF_STRING)
{
/* handle error */
}
Tokeniser interface provides functionality similar to strtok(3)
function. The tokeniser
is fully re-entrant with no hidden state, and its input string is not modified while being
parsed.
// declare and initialise tokeniser state
str_tok_state state;
str_tok_init(&state, source_string, delimiter_set);
// object to receive tokens
str token = str_null;
// token iterator
while(str_tok(&token, &state))
{
/* process "token" */
}
void str_tok_init(str_tok_state* const state, const str src, const str delim_set)
Initialises tokeniser state with the given source string and delimiter set. The delimiter set
is treated as bytes, not as UNICODE code points encoded in UTF-8.
bool str_tok(str* const dest, str_tok_state* const state)
Retrieves the next token and stores it in the dest
object. Returns true
if the token has
been read, or false
if the end of input has been reached. Retrieved token is always
a reference to a slice of the source string.
void str_tok_delim(str_tok_state* const state, const str delim_set)
Changes the delimiter set associated with the given tokeniser state. The delimiter set is
treated as bytes, not as UNICODE code points encoded in UTF-8.
All the tools are located in tools/
directory. Currently, there are the following tools:
-
file-to-str
: The script takes a file (text or binary) and a C variable name, and writes tostdout
C source code where the variable (of typestr
) is defined and initialised with the content of the file. -
gen-char-class
: Generates character classification functions that do the same as theirisw*()
counterparts under the current locale as specified byLC_ALL
environment variable. Runtools/gen-char-class --help
for further details, ortools/gen-char-class --space
to see an example of its output.
The library requires at least a C11 compiler. So far has been tested on Linux Mint versions
from 19.3 to 22.0, with gcc
versions from 9.5.0 to 13.2.0 (with either libc
or musl
),
and clang
versions up to 18.1.3; it is also reported to work on ALT Linux 9.1 for Elbrus, with
lcc
version 1.25.09.