• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1---
2layout: default
3title: UTF-8
4nav_order: 1
5parent: Chars and Strings
6---
7<!--
8© 2020 and later: Unicode, Inc. and others.
9License & terms of use: http://www.unicode.org/copyright.html
10-->
11
12# UTF-8
13
14*Note: This page is only relevant for C/C++. In Java, all strings are encoded in
15UTF-16, except for conversion from bytes to strings (via InputStreamReader or
16similar) and from strings to bytes (OutputStreamWriter etc.).*
17
18While most of ICU works with UTF-16 strings and uses data structures optimized
19for UTF-16, there are APIs that facilitate working with UTF-8, or are optimized
20for UTF-8, or work with Unicode code points (21-bit integer values) regardless
21of string encoding. Some data structures are designed to work equally well with
22UTF-16 and UTF-8.
23
24For UTF-8 strings, ICU normally uses `(const) char *` pointers and `int32_t`
25lengths, normally with semantics parallel to UTF-16 handling. (Input length=-1
26means NUL-terminated, output is NUL-terminated if there is space, output
27overflow is handled with preflighting; for details see the parent [Strings
28page](index.md).) Some newer APIs take an `icu::StringPiece` argument and write
29to an `icu::ByteSink` or to a string class object like `std::string`.
30
31## Conversion Between UTF-8 and UTF-16
32
33The simplest way to use UTF-8 strings in UTF-16 APIs is via the C++
34`icu::UnicodeString` methods `fromUTF8(const StringPiece &utf8)` and
35`toUTF8String(StringClass &result)`. There is also `toUTF8(ByteSink &sink)`.
36
37In C, `unicode/ustring.h` has functions like `u_strFromUTF8WithSub()` and
38`u_strToUTF8WithSub()`. (Also `u_strFromUTF8()`, `u_strToUTF8()` and
39`u_strFromUTF8Lenient()`.)
40
41The conversion functions in `unicode/ucnv.h` are intended for very flexible
42handling of conversion to/from external byte streams (with customizable error
43handling and support for split buffers at arbitrary boundaries) which is
44normally unnecessary for internal strings.
45
46Note: `icu::``UnicodeString` has constructors, `setTo()` and `extract()` methods
47which take either a converter object or a charset name. These can be used for
48UTF-8, but are not as efficient or convenient as the
49`fromUTF8()`/`toUTF8()`/`toUTF8String()` methods mentioned above. (Among
50conversion methods, APIs with a charset name are more convenient but internally
51open and close a converter; ones with a converter object parameter avoid this.)
52
53## UTF-8 as Default Charset
54
55ICU has many functions that take or return `char *` strings that are assumed to
56be in the default charset which should match the system encoding. Since this
57could be one of many charsets, and the charset can be different for different
58processes on the same system, ICU uses its conversion framework for converting
59to and from UTF-16.
60
61If it is known that the default charset is always UTF-8 on the target platform,
62then you should `#define`` U_CHARSET_IS_UTF8 1` in or before `unicode/utypes.h`.
63(For example, modify the default value there or pass `-D``U_CHARSET_IS_UTF8=1`
64as a compiler flag.) This will change most of the implementation code to use
65dedicated (simpler, faster) UTF-8 code paths and avoid dependencies on the
66conversion framework. (Avoiding such dependencies helps with statically linked
67libraries and may allow the use of `UCONFIG_NO_LEGACY_CONVERSION` or even
68`UCONFIG_NO_CONVERSION` \[see `unicode/uconfig.h`\].)
69
70## Low-Level UTF-8 String Operations
71
72`unicode/utf8.h` defines macros for UTF-8 with semantics parallel to the UTF-16
73macros in `unicode/utf16.h`. The macros handle many cases inline, but call
74internal functions for complicated parts of the UTF-8 encoding form. For
75example, the following code snippet counts white space characters in a string:
76
77```c
78#include "unicode/utypes.h"
79#include "unicode/stringpiece.h"
80#include "unicode/utf8.h"
81#include "unicode/uchar.h"
82
83int32_t countWhiteSpace(StringPiece sp) {
84    const char *s=sp.data();
85    int32_t length=sp.length();
86    int32_t count=0;
87    for(int32_t i=0; i<length;) {
88        UChar32 c;
89        U8_NEXT(s, i, length, c);
90        if(u_isUWhiteSpace(c)) {
91            ++count;
92        }
93    }
94    return count;
95}
96```
97
98## Dedicated UTF-8 APIs
99
100ICU has some APIs dedicated for UTF-8. They tend to have been added for "worker
101functions" like comparing strings, to avoid the string conversion overhead,
102rather than for "builder functions" like factory methods and attribute setters.
103
104For example, `icu::Collator::compareUTF8()` compares two UTF-8 strings
105incrementally, without converting all of the two strings to UTF-16 if there is
106an early base letter difference.
107
108`ucnv_convertEx()` can convert between UTF-8 and another charset, if one of the
109two `UConverter`s is a UTF-8 converter. The conversion *from UTF-8 to* most
110other charsets uses a dedicated, optimized code path, avoiding the pivot through
111UTF-16. (Conversion *from* other charsets *to UTF-8* could be optimized as well,
112but that has not been implemented yet as of ICU 4.4.)
113
114Other examples: (This list may or may not be complete.)
115
116*   ucasemap_utf8ToLower(), ucasemap_utf8ToUpper(), ucasemap_utf8ToTitle(),
117    ucasemap_utf8FoldCase()
118*   ucnvsel_selectForUTF8()
119*   icu::UnicodeSet::spanUTF8(), spanBackUTF8() and uset_spanUTF8(),
120    uset_spanBackUTF8() (These are highly optimized for UTF-8 processing.)
121*   ures_getUTF8String(), ures_getUTF8StringByIndex(), ures_getUTF8StringByKey()
122*   uspoof_checkUTF8(), uspoof_areConfusableUTF8(), uspoof_getSkeletonUTF8()
123
124## Abstract Text APIs
125
126ICU offers several interfaces for text access, designed for different use cases.
127(Some interfaces are simply newer and more modern than others.) Some ICU
128services work with some of these interfaces, and for some of these interfaces
129ICU offers UTF-8 implementations out of the box.
130
131`UText` can be used with `BreakIterator` APIs (character/word/sentence/...
132segmentation). `utext_openUTF8()` creates a read-only `UText` for a UTF-8
133string.
134
135*   *Note: In ICU 4.4 and before, BreakIterator only works with UTF-8 (or any
136    other charset with non-1:1 index conversion to UTF-16) if no dictionary is
137    supported. This excludes Thai word break. See [ticket #5532](https://unicode-org.atlassian.net/browse/ICU-5532).*
138*   *As a workaround for Thai word breaking, you can convert the string to
139    UTF-16 and convert indexes to UTF-8 string indexes via
140    `u_strToUTF8(dest=NULL, destCapacity=0, *destLength gets UTF-8 index).`*
141*   *ICU 4.4 has a technology preview for UText in the regular expression API,
142    but some of the UText regex API and semantics are likely to change for ICU
143    4.6. (Especially indexing semantics.)*
144
145A `UCharIterator` can be used with several collation APIs (although there is
146also the newer `icu::Collator::compareUTF8()`) and with `u_strCompareIter()`.
147`uiter_setUTF8()` creates a UCharIterator for a UTF-8 string.
148
149It is also possible to create a `CharacterIterator` subclass for UTF-8 strings,
150but `CharacterIterator` has a lot of virtual methods and it requires UTF-16
151string index semantics.
152