/*
- * Copyright 2013 Facebook, Inc.
+ * Copyright 2015 Facebook, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
#define FOLLY_BASE_STRING_H_
#include <exception>
+#include <stdarg.h>
#include <string>
#include <boost/type_traits.hpp>
-#ifdef __GNUC__
-# include <ext/hash_set>
-# include <ext/hash_map>
+#ifdef FOLLY_HAVE_DEPRECATED_ASSOC
+#ifdef _GLIBCXX_SYMVER
+#include <ext/hash_set>
+#include <ext/hash_map>
#endif
+#endif
+
+#include <unordered_set>
+#include <unordered_map>
-#include "folly/Conv.h"
-#include "folly/FBString.h"
-#include "folly/FBVector.h"
-#include "folly/Range.h"
-#include "folly/ScopeGuard.h"
+#include <folly/Conv.h>
+#include <folly/Demangle.h>
+#include <folly/FBString.h>
+#include <folly/FBVector.h>
+#include <folly/Portability.h>
+#include <folly/Range.h>
+#include <folly/ScopeGuard.h>
// Compatibility function, to make sure toStdString(s) can be called
// to convert a std::string or fbstring variable s into type std::string
return out;
}
+/**
+ * URI-escape a string. Appends the result to the output string.
+ *
+ * Alphanumeric characters and other characters marked as "unreserved" in RFC
+ * 3986 ( -_.~ ) are left unchanged. In PATH mode, the forward slash (/) is
+ * also left unchanged. In QUERY mode, spaces are replaced by '+'. All other
+ * characters are percent-encoded.
+ */
+enum class UriEscapeMode : unsigned char {
+ // The values are meaningful, see generate_escape_tables.py
+ ALL = 0,
+ QUERY = 1,
+ PATH = 2
+};
+template <class String>
+void uriEscape(StringPiece str,
+ String& out,
+ UriEscapeMode mode = UriEscapeMode::ALL);
+
+/**
+ * Similar to uriEscape above, but returns the escaped string.
+ */
+template <class String>
+String uriEscape(StringPiece str, UriEscapeMode mode = UriEscapeMode::ALL) {
+ String out;
+ uriEscape(str, out, mode);
+ return out;
+}
+
+/**
+ * URI-unescape a string. Appends the result to the output string.
+ *
+ * In QUERY mode, '+' are replaced by space. %XX sequences are decoded if
+ * XX is a valid hex sequence, otherwise we throw invalid_argument.
+ */
+template <class String>
+void uriUnescape(StringPiece str,
+ String& out,
+ UriEscapeMode mode = UriEscapeMode::ALL);
+
+/**
+ * Similar to uriUnescape above, but returns the unescaped string.
+ */
+template <class String>
+String uriUnescape(StringPiece str, UriEscapeMode mode = UriEscapeMode::ALL) {
+ String out;
+ uriUnescape(str, out, mode);
+ return out;
+}
+
/**
* stringPrintf is much like printf but deposits its result into a
* string. Two signatures are supported: the first simply returns the
* resulting string, and the second appends the produced characters to
* the specified string and returns a reference to it.
*/
-std::string stringPrintf(const char* format, ...)
- __attribute__ ((format (printf, 1, 2)));
+std::string stringPrintf(FOLLY_PRINTF_FORMAT const char* format, ...)
+ FOLLY_PRINTF_FORMAT_ATTR(1, 2);
-/** Similar to stringPrintf, with different signiture.
- */
-void stringPrintf(std::string* out, const char* fmt, ...)
- __attribute__ ((format (printf, 2, 3)));
+/* Similar to stringPrintf, with different signature. */
+void stringPrintf(std::string* out, FOLLY_PRINTF_FORMAT const char* fmt, ...)
+ FOLLY_PRINTF_FORMAT_ATTR(2, 3);
-std::string& stringAppendf(std::string* output, const char* format, ...)
- __attribute__ ((format (printf, 2, 3)));
+std::string& stringAppendf(std::string* output,
+ FOLLY_PRINTF_FORMAT const char* format, ...)
+ FOLLY_PRINTF_FORMAT_ATTR(2, 3);
+
+/**
+ * Similar to stringPrintf, but accepts a va_list argument.
+ *
+ * As with vsnprintf() itself, the value of ap is undefined after the call.
+ * These functions do not call va_end() on ap.
+ */
+std::string stringVPrintf(const char* format, va_list ap);
+void stringVPrintf(std::string* out, const char* format, va_list ap);
+std::string& stringVAppendf(std::string* out, const char* format, va_list ap);
/**
* Backslashify a string, that is, replace non-printable characters
* PRETTY_UNITS_METRIC - k, M, G, etc (goes up by 10^3 = 1000 each time)
* PRETTY_UNITS_BINARY - k, M, G, etc (goes up by 2^10 = 1024 each time)
* PRETTY_UNITS_BINARY_IEC - Ki, Mi, Gi, etc
- *
+ * PRETTY_SI - full SI metric prefixes from yocto to Yotta
+ * http://en.wikipedia.org/wiki/Metric_prefix
* @author Mark Rabkin <mrabkin@fb.com>
*/
enum PrettyType {
PRETTY_UNITS_BINARY,
PRETTY_UNITS_BINARY_IEC,
- PRETTY_NUM_TYPES
+ PRETTY_SI,
+ PRETTY_NUM_TYPES,
};
std::string prettyPrint(double val, PrettyType, bool addSpace = true);
+/**
+ * This utility converts StringPiece in pretty format (look above) to double,
+ * with progress information. Alters the StringPiece parameter
+ * to get rid of the already-parsed characters.
+ * Expects string in form <floating point number> {space}* [<suffix>]
+ * If string is not in correct format, utility finds longest valid prefix and
+ * if there at least one, returns double value based on that prefix and
+ * modifies string to what is left after parsing. Throws and std::range_error
+ * exception if there is no correct parse.
+ * Examples(for PRETTY_UNITS_METRIC):
+ * '10M' => 10 000 000
+ * '10 M' => 10 000 000
+ * '10' => 10
+ * '10 Mx' => 10 000 000, prettyString == "x"
+ * 'abc' => throws std::range_error
+ */
+double prettyToDouble(folly::StringPiece *const prettyString,
+ const PrettyType type);
+
+/*
+ * Same as prettyToDouble(folly::StringPiece*, PrettyType), but
+ * expects whole string to be correctly parseable. Throws std::range_error
+ * otherwise
+ */
+double prettyToDouble(folly::StringPiece prettyString, const PrettyType type);
+
/**
* Write a hex dump of size bytes starting at ptr to out.
*
fbstring errnoStr(int err);
/**
- * Return the demangled (prettyfied) version of a C++ type.
- *
- * This function tries to produce a human-readable type, but the type name will
- * be returned unchanged in case of error or if demangling isn't supported on
- * your system.
- *
- * Use for debugging -- do not rely on demangle() returning anything useful.
- *
- * This function may allocate memory (and therefore throw).
- */
-fbstring demangle(const char* name);
-inline fbstring demangle(const std::type_info& type) {
- return demangle(type.name());
-}
-
-/**
- * Debug string for an exception: include type and what().
+ * Debug string for an exception: include type and what(), if
+ * defined.
*/
inline fbstring exceptionStr(const std::exception& e) {
+#ifdef FOLLY_HAS_RTTI
return folly::to<fbstring>(demangle(typeid(e)), ": ", e.what());
+#else
+ return folly::to<fbstring>("Exception (no RTTI available): ", e.what());
+#endif
}
+// Empirically, this indicates if the runtime supports
+// std::exception_ptr, as not all (arm, for instance) do.
+#if defined(__GNUC__) && defined(__GCC_ATOMIC_INT_LOCK_FREE) && \
+ __GCC_ATOMIC_INT_LOCK_FREE > 1
inline fbstring exceptionStr(std::exception_ptr ep) {
try {
std::rethrow_exception(ep);
return "<unknown exception>";
}
}
+#endif
+
+template<typename E>
+auto exceptionStr(const E& e)
+ -> typename std::enable_if<!std::is_base_of<std::exception, E>::value,
+ fbstring>::type
+{
+#ifdef FOLLY_HAS_RTTI
+ return folly::to<fbstring>(demangle(typeid(e)));
+#else
+ return "Exception (no RTTI available)";
+#endif
+}
/*
* Split a string into a list of tokens by delimiter.
OutputIterator out,
bool ignoreEmpty = false);
+/*
+ * Split a string into a fixed number of string pieces and/or numeric types
+ * by delimiter. Any numeric type that folly::to<> can convert to from a
+ * string piece is supported as a target. Returns 'true' if the fields were
+ * all successfully populated. Returns 'false' if there were too few fields
+ * in the input, or too many fields if exact=true. Casting exceptions will
+ * not be caught.
+ *
+ * Examples:
+ *
+ * folly::StringPiece name, key, value;
+ * if (folly::split('\t', line, name, key, value))
+ * ...
+ *
+ * folly::StringPiece name;
+ * double value;
+ * int id;
+ * if (folly::split('\t', line, name, value, id))
+ * ...
+ *
+ * The 'exact' template parameter specifies how the function behaves when too
+ * many fields are present in the input string. When 'exact' is set to its
+ * default value of 'true', a call to split will fail if the number of fields in
+ * the input string does not exactly match the number of output parameters
+ * passed. If 'exact' is overridden to 'false', all remaining fields will be
+ * stored, unsplit, in the last field, as shown below:
+ *
+ * folly::StringPiece x, y.
+ * if (folly::split<false>(':', "a:b:c", x, y))
+ * assert(x == "a" && y == "b:c");
+ *
+ * Note that this will likely not work if the last field's target is of numeric
+ * type, in which case folly::to<> will throw an exception.
+ */
+template <class T>
+using IsSplitTargetType = std::integral_constant<bool,
+ std::is_arithmetic<T>::value ||
+ std::is_same<T, StringPiece>::value ||
+ IsSomeString<T>::value>;
+
+template<bool exact = true,
+ class Delim,
+ class OutputType,
+ class... OutputTypes>
+typename std::enable_if<IsSplitTargetType<OutputType>::value, bool>::type
+split(const Delim& delimiter,
+ StringPiece input,
+ OutputType& outHead,
+ OutputTypes&... outTail);
+
/*
* Join list of tokens.
*
join(delimiter, container.begin(), container.end(), output);
}
+template <class Delim, class Value, class String>
+void join(const Delim& delimiter,
+ const std::initializer_list<Value>& values,
+ String& output) {
+ join(delimiter, values.begin(), values.end(), output);
+}
+
template <class Delim, class Container>
std::string join(const Delim& delimiter,
const Container& container) {
return output;
}
-} // namespace folly
+template <class Delim, class Value>
+std::string join(const Delim& delimiter,
+ const std::initializer_list<Value>& values) {
+ std::string output;
+ join(delimiter, values.begin(), values.end(), output);
+ return output;
+}
+
+template <class Delim,
+ class Iterator,
+ typename std::enable_if<std::is_same<
+ typename std::iterator_traits<Iterator>::iterator_category,
+ std::random_access_iterator_tag>::value>::type* = nullptr>
+std::string join(const Delim& delimiter, Iterator begin, Iterator end) {
+ std::string output;
+ join(delimiter, begin, end, output);
+ return output;
+}
-// Hash functions for string and fbstring usable with e.g. hash_map
-#ifdef __GNUC__
-namespace __gnu_cxx {
+/**
+ * Returns a subpiece with all whitespace removed from the front of @sp.
+ * Whitespace means any of [' ', '\n', '\r', '\t'].
+ */
+StringPiece ltrimWhitespace(StringPiece sp);
-template <class C>
-struct hash<folly::basic_fbstring<C> > : private hash<const C*> {
- size_t operator()(const folly::basic_fbstring<C> & s) const {
- return hash<const C*>::operator()(s.c_str());
- }
-};
+/**
+ * Returns a subpiece with all whitespace removed from the back of @sp.
+ * Whitespace means any of [' ', '\n', '\r', '\t'].
+ */
+StringPiece rtrimWhitespace(StringPiece sp);
-template <class C>
-struct hash<std::basic_string<C> > : private hash<const C*> {
- size_t operator()(const std::basic_string<C> & s) const {
- return hash<const C*>::operator()(s.c_str());
- }
-};
+/**
+ * Returns a subpiece with all whitespace removed from the back and front of @sp.
+ * Whitespace means any of [' ', '\n', '\r', '\t'].
+ */
+inline StringPiece trimWhitespace(StringPiece sp) {
+ return ltrimWhitespace(rtrimWhitespace(sp));
+}
-} // namespace __gnu_cxx
-#endif
+/**
+ * Returns a subpiece with all whitespace removed from the front of @sp.
+ * Whitespace means any of [' ', '\n', '\r', '\t'].
+ * DEPRECATED: @see ltrimWhitespace @see rtrimWhitespace
+ */
+inline StringPiece skipWhitespace(StringPiece sp) {
+ return ltrimWhitespace(sp);
+}
+
+/**
+ * Fast, in-place lowercasing of ASCII alphabetic characters in strings.
+ * Leaves all other characters unchanged, including those with the 0x80
+ * bit set.
+ * @param str String to convert
+ * @param len Length of str, in bytes
+ */
+void toLowerAscii(char* str, size_t length);
+
+inline void toLowerAscii(MutableStringPiece str) {
+ toLowerAscii(str.begin(), str.size());
+}
+
+} // namespace folly
// Hook into boost's type traits
namespace boost {
};
} // namespace boost
-#include "folly/String-inl.h"
+#include <folly/String-inl.h>
#endif