2 * Copyright 2014 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_
22 #include <boost/type_traits.hpp>
24 #ifdef _GLIBCXX_SYMVER
25 #include <ext/hash_set>
26 #include <ext/hash_map>
29 #include <unordered_set>
30 #include <unordered_map>
32 #include "folly/Conv.h"
33 #include "folly/Demangle.h"
34 #include "folly/FBString.h"
35 #include "folly/FBVector.h"
36 #include "folly/Portability.h"
37 #include "folly/Range.h"
38 #include "folly/ScopeGuard.h"
40 // Compatibility function, to make sure toStdString(s) can be called
41 // to convert a std::string or fbstring variable s into type std::string
42 // with very little overhead if s was already std::string
46 std::string toStdString(const folly::fbstring& s) {
47 return std::string(s.data(), s.size());
51 const std::string& toStdString(const std::string& s) {
55 // If called with a temporary, the compiler will select this overload instead
56 // of the above, so we don't return a (lvalue) reference to a temporary.
58 std::string&& toStdString(std::string&& s) {
63 * C-Escape a string, making it suitable for representation as a C string
64 * literal. Appends the result to the output string.
66 * Backslashes all occurrences of backslash and double-quote:
70 * Replaces all non-printable ASCII characters with backslash-octal
74 * Note that we use backslash-octal instead of backslash-hex because the octal
75 * representation is guaranteed to consume no more than 3 characters; "\3760"
76 * represents two characters, one with value 254, and one with value 48 ('0'),
77 * whereas "\xfe0" represents only one character (with value 4064, which leads
78 * to implementation-defined behavior).
80 template <class String>
81 void cEscape(StringPiece str, String& out);
84 * Similar to cEscape above, but returns the escaped string.
86 template <class String>
87 String cEscape(StringPiece str) {
94 * C-Unescape a string; the opposite of cEscape above. Appends the result
95 * to the output string.
97 * Recognizes the standard C escape sequences:
99 * \' \" \? \\ \a \b \f \n \r \t \v
103 * In strict mode (default), throws std::invalid_argument if it encounters
104 * an unrecognized escape sequence. In non-strict mode, it leaves
105 * the escape sequence unchanged.
107 template <class String>
108 void cUnescape(StringPiece str, String& out, bool strict = true);
111 * Similar to cUnescape above, but returns the escaped string.
113 template <class String>
114 String cUnescape(StringPiece str, bool strict = true) {
116 cUnescape(str, out, strict);
121 * URI-escape a string. Appends the result to the output string.
123 * Alphanumeric characters and other characters marked as "unreserved" in RFC
124 * 3986 ( -_.~ ) are left unchanged. In PATH mode, the forward slash (/) is
125 * also left unchanged. In QUERY mode, spaces are replaced by '+'. All other
126 * characters are percent-encoded.
128 enum class UriEscapeMode : unsigned char {
129 // The values are meaningful, see generate_escape_tables.py
134 template <class String>
135 void uriEscape(StringPiece str,
137 UriEscapeMode mode = UriEscapeMode::ALL);
140 * Similar to uriEscape above, but returns the escaped string.
142 template <class String>
143 String uriEscape(StringPiece str, UriEscapeMode mode = UriEscapeMode::ALL) {
145 uriEscape(str, out, mode);
150 * URI-unescape a string. Appends the result to the output string.
152 * In QUERY mode, '+' are replaced by space. %XX sequences are decoded if
153 * XX is a valid hex sequence, otherwise we throw invalid_argument.
155 template <class String>
156 void uriUnescape(StringPiece str,
158 UriEscapeMode mode = UriEscapeMode::ALL);
161 * Similar to uriUnescape above, but returns the unescaped string.
163 template <class String>
164 String uriUnescape(StringPiece str, UriEscapeMode mode = UriEscapeMode::ALL) {
166 uriUnescape(str, out, mode);
171 * stringPrintf is much like printf but deposits its result into a
172 * string. Two signatures are supported: the first simply returns the
173 * resulting string, and the second appends the produced characters to
174 * the specified string and returns a reference to it.
176 std::string stringPrintf(const char* format, ...)
177 __attribute__ ((format (printf, 1, 2)));
179 /** Similar to stringPrintf, with different signiture.
181 void stringPrintf(std::string* out, const char* fmt, ...)
182 __attribute__ ((format (printf, 2, 3)));
184 std::string& stringAppendf(std::string* output, const char* format, ...)
185 __attribute__ ((format (printf, 2, 3)));
188 * Backslashify a string, that is, replace non-printable characters
189 * with C-style (but NOT C compliant) "\xHH" encoding. If hex_style
190 * is false, then shorthand notations like "\0" will be used instead
191 * of "\x00" for the most common backslash cases.
193 * There are two forms, one returning the input string, and one
194 * creating output in the specified output string.
196 * This is mainly intended for printing to a terminal, so it is not
197 * particularly optimized.
199 * Do *not* use this in situations where you expect to be able to feed
200 * the string to a C or C++ compiler, as there are nuances with how C
201 * parses such strings that lead to failures. This is for display
202 * purposed only. If you want a string you can embed for use in C or
203 * C++, use cEscape instead. This function is for display purposes
206 template <class String1, class String2>
207 void backslashify(const String1& input, String2& output, bool hex_style=false);
209 template <class String>
210 String backslashify(const String& input, bool hex_style=false) {
212 backslashify(input, output, hex_style);
217 * Take a string and "humanify" it -- that is, make it look better.
218 * Since "better" is subjective, caveat emptor. The basic approach is
219 * to count the number of unprintable characters. If there are none,
220 * then the output is the input. If there are relatively few, or if
221 * there is a long "enough" prefix of printable characters, use
222 * backslashify. If it is mostly binary, then simply hex encode.
224 * This is an attempt to make a computer smart, and so likely is wrong
227 template <class String1, class String2>
228 void humanify(const String1& input, String2& output);
230 template <class String>
231 String humanify(const String& input) {
233 humanify(input, output);
238 * Same functionality as Python's binascii.hexlify. Returns true
239 * on successful conversion.
241 * If append_output is true, append data to the output rather than
244 template<class InputString, class OutputString>
245 bool hexlify(const InputString& input, OutputString& output,
249 * Same functionality as Python's binascii.unhexlify. Returns true
250 * on successful conversion.
252 template<class InputString, class OutputString>
253 bool unhexlify(const InputString& input, OutputString& output);
256 * A pretty-printer for numbers that appends suffixes of units of the
257 * given type. It prints 4 sig-figs of value with the most
260 * If `addSpace' is true, we put a space between the units suffix and
264 * PRETTY_TIME - s, ms, us, ns, etc.
265 * PRETTY_BYTES_METRIC - kB, MB, GB, etc (goes up by 10^3 = 1000 each time)
266 * PRETTY_BYTES - kB, MB, GB, etc (goes up by 2^10 = 1024 each time)
267 * PRETTY_BYTES_IEC - KiB, MiB, GiB, etc
268 * PRETTY_UNITS_METRIC - k, M, G, etc (goes up by 10^3 = 1000 each time)
269 * PRETTY_UNITS_BINARY - k, M, G, etc (goes up by 2^10 = 1024 each time)
270 * PRETTY_UNITS_BINARY_IEC - Ki, Mi, Gi, etc
271 * PRETTY_SI - full SI metric prefixes from yocto to Yotta
272 * http://en.wikipedia.org/wiki/Metric_prefix
273 * @author Mark Rabkin <mrabkin@fb.com>
280 PRETTY_BYTES = PRETTY_BYTES_BINARY,
281 PRETTY_BYTES_BINARY_IEC,
282 PRETTY_BYTES_IEC = PRETTY_BYTES_BINARY_IEC,
286 PRETTY_UNITS_BINARY_IEC,
292 std::string prettyPrint(double val, PrettyType, bool addSpace = true);
295 * This utility converts StringPiece in pretty format (look above) to double,
296 * with progress information. Alters the StringPiece parameter
297 * to get rid of the already-parsed characters.
298 * Expects string in form <floating point number> {space}* [<suffix>]
299 * If string is not in correct format, utility finds longest valid prefix and
300 * if there at least one, returns double value based on that prefix and
301 * modifies string to what is left after parsing. Throws and std::range_error
302 * exception if there is no correct parse.
303 * Examples(for PRETTY_UNITS_METRIC):
304 * '10M' => 10 000 000
305 * '10 M' => 10 000 000
307 * '10 Mx' => 10 000 000, prettyString == "x"
308 * 'abc' => throws std::range_error
310 double prettyToDouble(folly::StringPiece *const prettyString,
311 const PrettyType type);
314 * Same as prettyToDouble(folly::StringPiece*, PrettyType), but
315 * expects whole string to be correctly parseable. Throws std::range_error
318 double prettyToDouble(folly::StringPiece prettyString, const PrettyType type);
321 * Write a hex dump of size bytes starting at ptr to out.
323 * The hex dump is formatted as follows:
325 * for the string "abcdefghijklmnopqrstuvwxyz\x02"
326 00000000 61 62 63 64 65 66 67 68 69 6a 6b 6c 6d 6e 6f 70 |abcdefghijklmnop|
327 00000010 71 72 73 74 75 76 77 78 79 7a 02 |qrstuvwxyz. |
329 * that is, we write 16 bytes per line, both as hex bytes and as printable
330 * characters. Non-printable characters are replaced with '.'
331 * Lines are written to out one by one (one StringPiece at a time) without
334 template <class OutIt>
335 void hexDump(const void* ptr, size_t size, OutIt out);
338 * Return the hex dump of size bytes starting at ptr as a string.
340 std::string hexDump(const void* ptr, size_t size);
343 * Return a fbstring containing the description of the given errno value.
344 * Takes care not to overwrite the actual system errno, so calling
345 * errnoStr(errno) is valid.
347 fbstring errnoStr(int err);
350 * Debug string for an exception: include type and what().
352 inline fbstring exceptionStr(const std::exception& e) {
353 return folly::to<fbstring>(demangle(typeid(e)), ": ", e.what());
356 inline fbstring exceptionStr(std::exception_ptr ep) {
358 std::rethrow_exception(ep);
359 } catch (const std::exception& e) {
360 return exceptionStr(e);
362 return "<unknown exception>";
367 * Split a string into a list of tokens by delimiter.
369 * The split interface here supports different output types, selected
370 * at compile time: StringPiece, fbstring, or std::string. If you are
371 * using a vector to hold the output, it detects the type based on
372 * what your vector contains. If the output vector is not empty, split
373 * will append to the end of the vector.
375 * You can also use splitTo() to write the output to an arbitrary
376 * OutputIterator (e.g. std::inserter() on a std::set<>), in which
377 * case you have to tell the function the type. (Rationale:
378 * OutputIterators don't have a value_type, so we can't detect the
379 * type in splitTo without being told.)
383 * std::vector<folly::StringPiece> v;
384 * folly::split(":", "asd:bsd", v);
386 * std::set<StringPiece> s;
387 * folly::splitTo<StringPiece>(":", "asd:bsd:asd:csd",
388 * std::inserter(s, s.begin()));
390 * Split also takes a flag (ignoreEmpty) that indicates whether adjacent
391 * delimiters should be treated as one single separator (ignoring empty tokens)
392 * or not (generating empty tokens).
395 template<class Delim, class String, class OutputType>
396 void split(const Delim& delimiter,
398 std::vector<OutputType>& out,
399 bool ignoreEmpty = false);
401 template<class Delim, class String, class OutputType>
402 void split(const Delim& delimiter,
404 folly::fbvector<OutputType>& out,
405 bool ignoreEmpty = false);
407 template<class OutputValueType, class Delim, class String,
408 class OutputIterator>
409 void splitTo(const Delim& delimiter,
412 bool ignoreEmpty = false);
415 * Split a string into a fixed number of string pieces and/or numeric types
416 * by delimiter. Any numeric type that folly::to<> can convert to from a
417 * string piece is supported as a target. Returns 'true' if the fields were
418 * all successfully populated.
422 * folly::StringPiece name, key, value;
423 * if (folly::split('\t', line, name, key, value))
426 * folly::StringPiece name;
429 * if (folly::split('\t', line, name, value, id))
432 * The 'exact' template parameter specifies how the function behaves when too
433 * many fields are present in the input string. When 'exact' is set to its
434 * default value of 'true', a call to split will fail if the number of fields in
435 * the input string does not exactly match the number of output parameters
436 * passed. If 'exact' is overridden to 'false', all remaining fields will be
437 * stored, unsplit, in the last field, as shown below:
439 * folly::StringPiece x, y.
440 * if (folly::split<false>(':', "a:b:c", x, y))
441 * assert(x == "a" && y == "b:c");
443 * Note that this will likely not work if the last field's target is of numeric
444 * type, in which case folly::to<> will throw an exception.
447 using IsSplitTargetType = std::integral_constant<bool,
448 std::is_arithmetic<T>::value ||
449 std::is_same<T, StringPiece>::value>;
451 template<bool exact = true,
454 class... OutputTypes>
455 typename std::enable_if<IsSplitTargetType<OutputType>::value, bool>::type
456 split(const Delim& delimiter,
459 OutputTypes&... outTail);
462 * Join list of tokens.
464 * Stores a string representation of tokens in the same order with
465 * deliminer between each element.
468 template <class Delim, class Iterator, class String>
469 void join(const Delim& delimiter,
474 template <class Delim, class Container, class String>
475 void join(const Delim& delimiter,
476 const Container& container,
478 join(delimiter, container.begin(), container.end(), output);
481 template <class Delim, class Value, class String>
482 void join(const Delim& delimiter,
483 const std::initializer_list<Value>& values,
485 join(delimiter, values.begin(), values.end(), output);
488 template <class Delim, class Container>
489 std::string join(const Delim& delimiter,
490 const Container& container) {
492 join(delimiter, container.begin(), container.end(), output);
496 template <class Delim, class Value>
497 std::string join(const Delim& delimiter,
498 const std::initializer_list<Value>& values) {
500 join(delimiter, values.begin(), values.end(), output);
506 // Hash functions to make std::string usable with e.g. hash_map
508 // Handle interaction with different C++ standard libraries, which
509 // expect these types to be in different namespaces.
513 struct hash<std::basic_string<C> > : private hash<const C*> {
514 size_t operator()(const std::basic_string<C> & s) const {
515 return hash<const C*>::operator()(s.c_str());
521 #if defined(_GLIBCXX_SYMVER) && !defined(__BIONIC__)
522 namespace __gnu_cxx {
525 struct hash<std::basic_string<C> > : private hash<const C*> {
526 size_t operator()(const std::basic_string<C> & s) const {
527 return hash<const C*>::operator()(s.c_str());
534 // Hook into boost's type traits
537 struct has_nothrow_constructor<folly::basic_fbstring<T> > : true_type {
538 enum { value = true };
542 #include "folly/String-inl.h"