2 * Copyright 2015 Facebook, Inc.
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
8 * http://www.apache.org/licenses/LICENSE-2.0
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
17 #ifndef FOLLY_BASE_STRING_H_
18 #define FOLLY_BASE_STRING_H_
23 #include <boost/type_traits.hpp>
25 #ifdef FOLLY_HAVE_DEPRECATED_ASSOC
26 #ifdef _GLIBCXX_SYMVER
27 #include <ext/hash_set>
28 #include <ext/hash_map>
32 #include <unordered_set>
33 #include <unordered_map>
35 #include <folly/Conv.h>
36 #include <folly/Demangle.h>
37 #include <folly/FBString.h>
38 #include <folly/FBVector.h>
39 #include <folly/Portability.h>
40 #include <folly/Range.h>
41 #include <folly/ScopeGuard.h>
43 // Compatibility function, to make sure toStdString(s) can be called
44 // to convert a std::string or fbstring variable s into type std::string
45 // with very little overhead if s was already std::string
49 std::string toStdString(const folly::fbstring& s) {
50 return std::string(s.data(), s.size());
54 const std::string& toStdString(const std::string& s) {
58 // If called with a temporary, the compiler will select this overload instead
59 // of the above, so we don't return a (lvalue) reference to a temporary.
61 std::string&& toStdString(std::string&& s) {
66 * C-Escape a string, making it suitable for representation as a C string
67 * literal. Appends the result to the output string.
69 * Backslashes all occurrences of backslash and double-quote:
73 * Replaces all non-printable ASCII characters with backslash-octal
77 * Note that we use backslash-octal instead of backslash-hex because the octal
78 * representation is guaranteed to consume no more than 3 characters; "\3760"
79 * represents two characters, one with value 254, and one with value 48 ('0'),
80 * whereas "\xfe0" represents only one character (with value 4064, which leads
81 * to implementation-defined behavior).
83 template <class String>
84 void cEscape(StringPiece str, String& out);
87 * Similar to cEscape above, but returns the escaped string.
89 template <class String>
90 String cEscape(StringPiece str) {
97 * C-Unescape a string; the opposite of cEscape above. Appends the result
98 * to the output string.
100 * Recognizes the standard C escape sequences:
102 * \' \" \? \\ \a \b \f \n \r \t \v
106 * In strict mode (default), throws std::invalid_argument if it encounters
107 * an unrecognized escape sequence. In non-strict mode, it leaves
108 * the escape sequence unchanged.
110 template <class String>
111 void cUnescape(StringPiece str, String& out, bool strict = true);
114 * Similar to cUnescape above, but returns the escaped string.
116 template <class String>
117 String cUnescape(StringPiece str, bool strict = true) {
119 cUnescape(str, out, strict);
124 * URI-escape a string. Appends the result to the output string.
126 * Alphanumeric characters and other characters marked as "unreserved" in RFC
127 * 3986 ( -_.~ ) are left unchanged. In PATH mode, the forward slash (/) is
128 * also left unchanged. In QUERY mode, spaces are replaced by '+'. All other
129 * characters are percent-encoded.
131 enum class UriEscapeMode : unsigned char {
132 // The values are meaningful, see generate_escape_tables.py
137 template <class String>
138 void uriEscape(StringPiece str,
140 UriEscapeMode mode = UriEscapeMode::ALL);
143 * Similar to uriEscape above, but returns the escaped string.
145 template <class String>
146 String uriEscape(StringPiece str, UriEscapeMode mode = UriEscapeMode::ALL) {
148 uriEscape(str, out, mode);
153 * URI-unescape a string. Appends the result to the output string.
155 * In QUERY mode, '+' are replaced by space. %XX sequences are decoded if
156 * XX is a valid hex sequence, otherwise we throw invalid_argument.
158 template <class String>
159 void uriUnescape(StringPiece str,
161 UriEscapeMode mode = UriEscapeMode::ALL);
164 * Similar to uriUnescape above, but returns the unescaped string.
166 template <class String>
167 String uriUnescape(StringPiece str, UriEscapeMode mode = UriEscapeMode::ALL) {
169 uriUnescape(str, out, mode);
174 * stringPrintf is much like printf but deposits its result into a
175 * string. Two signatures are supported: the first simply returns the
176 * resulting string, and the second appends the produced characters to
177 * the specified string and returns a reference to it.
179 std::string stringPrintf(FOLLY_PRINTF_FORMAT const char* format, ...)
180 FOLLY_PRINTF_FORMAT_ATTR(1, 2);
182 /* Similar to stringPrintf, with different signature. */
183 void stringPrintf(std::string* out, FOLLY_PRINTF_FORMAT const char* fmt, ...)
184 FOLLY_PRINTF_FORMAT_ATTR(2, 3);
186 std::string& stringAppendf(std::string* output,
187 FOLLY_PRINTF_FORMAT const char* format, ...)
188 FOLLY_PRINTF_FORMAT_ATTR(2, 3);
191 * Similar to stringPrintf, but accepts a va_list argument.
193 * As with vsnprintf() itself, the value of ap is undefined after the call.
194 * These functions do not call va_end() on ap.
196 std::string stringVPrintf(const char* format, va_list ap);
197 void stringVPrintf(std::string* out, const char* format, va_list ap);
198 std::string& stringVAppendf(std::string* out, const char* format, va_list ap);
201 * Backslashify a string, that is, replace non-printable characters
202 * with C-style (but NOT C compliant) "\xHH" encoding. If hex_style
203 * is false, then shorthand notations like "\0" will be used instead
204 * of "\x00" for the most common backslash cases.
206 * There are two forms, one returning the input string, and one
207 * creating output in the specified output string.
209 * This is mainly intended for printing to a terminal, so it is not
210 * particularly optimized.
212 * Do *not* use this in situations where you expect to be able to feed
213 * the string to a C or C++ compiler, as there are nuances with how C
214 * parses such strings that lead to failures. This is for display
215 * purposed only. If you want a string you can embed for use in C or
216 * C++, use cEscape instead. This function is for display purposes
219 template <class String1, class String2>
220 void backslashify(const String1& input, String2& output, bool hex_style=false);
222 template <class String>
223 String backslashify(const String& input, bool hex_style=false) {
225 backslashify(input, output, hex_style);
230 * Take a string and "humanify" it -- that is, make it look better.
231 * Since "better" is subjective, caveat emptor. The basic approach is
232 * to count the number of unprintable characters. If there are none,
233 * then the output is the input. If there are relatively few, or if
234 * there is a long "enough" prefix of printable characters, use
235 * backslashify. If it is mostly binary, then simply hex encode.
237 * This is an attempt to make a computer smart, and so likely is wrong
240 template <class String1, class String2>
241 void humanify(const String1& input, String2& output);
243 template <class String>
244 String humanify(const String& input) {
246 humanify(input, output);
251 * Same functionality as Python's binascii.hexlify. Returns true
252 * on successful conversion.
254 * If append_output is true, append data to the output rather than
257 template<class InputString, class OutputString>
258 bool hexlify(const InputString& input, OutputString& output,
262 * Same functionality as Python's binascii.unhexlify. Returns true
263 * on successful conversion.
265 template<class InputString, class OutputString>
266 bool unhexlify(const InputString& input, OutputString& output);
269 * A pretty-printer for numbers that appends suffixes of units of the
270 * given type. It prints 4 sig-figs of value with the most
273 * If `addSpace' is true, we put a space between the units suffix and
277 * PRETTY_TIME - s, ms, us, ns, etc.
278 * PRETTY_BYTES_METRIC - kB, MB, GB, etc (goes up by 10^3 = 1000 each time)
279 * PRETTY_BYTES - kB, MB, GB, etc (goes up by 2^10 = 1024 each time)
280 * PRETTY_BYTES_IEC - KiB, MiB, GiB, etc
281 * PRETTY_UNITS_METRIC - k, M, G, etc (goes up by 10^3 = 1000 each time)
282 * PRETTY_UNITS_BINARY - k, M, G, etc (goes up by 2^10 = 1024 each time)
283 * PRETTY_UNITS_BINARY_IEC - Ki, Mi, Gi, etc
284 * PRETTY_SI - full SI metric prefixes from yocto to Yotta
285 * http://en.wikipedia.org/wiki/Metric_prefix
286 * @author Mark Rabkin <mrabkin@fb.com>
293 PRETTY_BYTES = PRETTY_BYTES_BINARY,
294 PRETTY_BYTES_BINARY_IEC,
295 PRETTY_BYTES_IEC = PRETTY_BYTES_BINARY_IEC,
299 PRETTY_UNITS_BINARY_IEC,
305 std::string prettyPrint(double val, PrettyType, bool addSpace = true);
308 * This utility converts StringPiece in pretty format (look above) to double,
309 * with progress information. Alters the StringPiece parameter
310 * to get rid of the already-parsed characters.
311 * Expects string in form <floating point number> {space}* [<suffix>]
312 * If string is not in correct format, utility finds longest valid prefix and
313 * if there at least one, returns double value based on that prefix and
314 * modifies string to what is left after parsing. Throws and std::range_error
315 * exception if there is no correct parse.
316 * Examples(for PRETTY_UNITS_METRIC):
317 * '10M' => 10 000 000
318 * '10 M' => 10 000 000
320 * '10 Mx' => 10 000 000, prettyString == "x"
321 * 'abc' => throws std::range_error
323 double prettyToDouble(folly::StringPiece *const prettyString,
324 const PrettyType type);
327 * Same as prettyToDouble(folly::StringPiece*, PrettyType), but
328 * expects whole string to be correctly parseable. Throws std::range_error
331 double prettyToDouble(folly::StringPiece prettyString, const PrettyType type);
334 * Write a hex dump of size bytes starting at ptr to out.
336 * The hex dump is formatted as follows:
338 * for the string "abcdefghijklmnopqrstuvwxyz\x02"
339 00000000 61 62 63 64 65 66 67 68 69 6a 6b 6c 6d 6e 6f 70 |abcdefghijklmnop|
340 00000010 71 72 73 74 75 76 77 78 79 7a 02 |qrstuvwxyz. |
342 * that is, we write 16 bytes per line, both as hex bytes and as printable
343 * characters. Non-printable characters are replaced with '.'
344 * Lines are written to out one by one (one StringPiece at a time) without
347 template <class OutIt>
348 void hexDump(const void* ptr, size_t size, OutIt out);
351 * Return the hex dump of size bytes starting at ptr as a string.
353 std::string hexDump(const void* ptr, size_t size);
356 * Return a fbstring containing the description of the given errno value.
357 * Takes care not to overwrite the actual system errno, so calling
358 * errnoStr(errno) is valid.
360 fbstring errnoStr(int err);
363 * Debug string for an exception: include type and what(), if
366 inline fbstring exceptionStr(const std::exception& e) {
367 return folly::to<fbstring>(demangle(typeid(e)), ": ", e.what());
370 inline fbstring exceptionStr(std::exception_ptr ep) {
372 std::rethrow_exception(ep);
373 } catch (const std::exception& e) {
374 return exceptionStr(e);
376 return "<unknown exception>";
381 auto exceptionStr(const E& e)
382 -> typename std::enable_if<!std::is_base_of<std::exception, E>::value,
385 return folly::to<fbstring>(demangle(typeid(e)));
389 * Split a string into a list of tokens by delimiter.
391 * The split interface here supports different output types, selected
392 * at compile time: StringPiece, fbstring, or std::string. If you are
393 * using a vector to hold the output, it detects the type based on
394 * what your vector contains. If the output vector is not empty, split
395 * will append to the end of the vector.
397 * You can also use splitTo() to write the output to an arbitrary
398 * OutputIterator (e.g. std::inserter() on a std::set<>), in which
399 * case you have to tell the function the type. (Rationale:
400 * OutputIterators don't have a value_type, so we can't detect the
401 * type in splitTo without being told.)
405 * std::vector<folly::StringPiece> v;
406 * folly::split(":", "asd:bsd", v);
408 * std::set<StringPiece> s;
409 * folly::splitTo<StringPiece>(":", "asd:bsd:asd:csd",
410 * std::inserter(s, s.begin()));
412 * Split also takes a flag (ignoreEmpty) that indicates whether adjacent
413 * delimiters should be treated as one single separator (ignoring empty tokens)
414 * or not (generating empty tokens).
417 template<class Delim, class String, class OutputType>
418 void split(const Delim& delimiter,
420 std::vector<OutputType>& out,
421 bool ignoreEmpty = false);
423 template<class Delim, class String, class OutputType>
424 void split(const Delim& delimiter,
426 folly::fbvector<OutputType>& out,
427 bool ignoreEmpty = false);
429 template<class OutputValueType, class Delim, class String,
430 class OutputIterator>
431 void splitTo(const Delim& delimiter,
434 bool ignoreEmpty = false);
437 * Split a string into a fixed number of string pieces and/or numeric types
438 * by delimiter. Any numeric type that folly::to<> can convert to from a
439 * string piece is supported as a target. Returns 'true' if the fields were
440 * all successfully populated. Returns 'false' if there were too few fields
441 * in the input, or too many fields if exact=true. Casting exceptions will
446 * folly::StringPiece name, key, value;
447 * if (folly::split('\t', line, name, key, value))
450 * folly::StringPiece name;
453 * if (folly::split('\t', line, name, value, id))
456 * The 'exact' template parameter specifies how the function behaves when too
457 * many fields are present in the input string. When 'exact' is set to its
458 * default value of 'true', a call to split will fail if the number of fields in
459 * the input string does not exactly match the number of output parameters
460 * passed. If 'exact' is overridden to 'false', all remaining fields will be
461 * stored, unsplit, in the last field, as shown below:
463 * folly::StringPiece x, y.
464 * if (folly::split<false>(':', "a:b:c", x, y))
465 * assert(x == "a" && y == "b:c");
467 * Note that this will likely not work if the last field's target is of numeric
468 * type, in which case folly::to<> will throw an exception.
471 using IsSplitTargetType = std::integral_constant<bool,
472 std::is_arithmetic<T>::value ||
473 std::is_same<T, StringPiece>::value ||
474 IsSomeString<T>::value>;
476 template<bool exact = true,
479 class... OutputTypes>
480 typename std::enable_if<IsSplitTargetType<OutputType>::value, bool>::type
481 split(const Delim& delimiter,
484 OutputTypes&... outTail);
487 * Join list of tokens.
489 * Stores a string representation of tokens in the same order with
490 * deliminer between each element.
493 template <class Delim, class Iterator, class String>
494 void join(const Delim& delimiter,
499 template <class Delim, class Container, class String>
500 void join(const Delim& delimiter,
501 const Container& container,
503 join(delimiter, container.begin(), container.end(), output);
506 template <class Delim, class Value, class String>
507 void join(const Delim& delimiter,
508 const std::initializer_list<Value>& values,
510 join(delimiter, values.begin(), values.end(), output);
513 template <class Delim, class Container>
514 std::string join(const Delim& delimiter,
515 const Container& container) {
517 join(delimiter, container.begin(), container.end(), output);
521 template <class Delim, class Value>
522 std::string join(const Delim& delimiter,
523 const std::initializer_list<Value>& values) {
525 join(delimiter, values.begin(), values.end(), output);
529 template <class Delim,
531 typename std::enable_if<std::is_same<
532 typename std::iterator_traits<Iterator>::iterator_category,
533 std::random_access_iterator_tag>::value>::type* = nullptr>
534 std::string join(const Delim& delimiter, Iterator begin, Iterator end) {
536 join(delimiter, begin, end, output);
541 * Returns a subpiece with all whitespace removed from the front of @sp.
542 * Whitespace means any of [' ', '\n', '\r', '\t'].
544 StringPiece skipWhitespace(StringPiece sp);
547 * Fast, in-place lowercasing of ASCII alphabetic characters in strings.
548 * Leaves all other characters unchanged, including those with the 0x80
550 * @param str String to convert
551 * @param len Length of str, in bytes
553 void toLowerAscii(char* str, size_t length);
555 inline void toLowerAscii(MutableStringPiece str) {
556 toLowerAscii(str.begin(), str.size());
561 // Hook into boost's type traits
564 struct has_nothrow_constructor<folly::basic_fbstring<T> > : true_type {
565 enum { value = true };
569 #include <folly/String-inl.h>