change the mechanism for "internal" buffer storage

[folly.git] / folly / String.h

diff --git a/folly/String.h b/folly/String.h
index c843eed8b1b5d4768570fc9dd0aa528092e2858e..da93cd265d106a296e60a651aad59cd1cfb16bfd 100644 (file)
--- a/folly/String.h
+++ b/folly/String.h
@@ -112,6 +112,56 @@ String cUnescape(StringPiece str, bool strict = true) {
   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
@@ -272,13 +322,36 @@ fbstring errnoStr(int err);
  *
  * Use for debugging -- do not rely on demangle() returning anything useful.
  *
- * This function may allocate memory (and therefore throw).
+ * This function may allocate memory (and therefore throw std::bad_alloc).
  */
 fbstring demangle(const char* name);
 inline fbstring demangle(const std::type_info& type) {
   return demangle(type.name());
 }
 
+/**
+ * Return the demangled (prettyfied) version of a C++ type in a user-provided
+ * buffer.
+ *
+ * The semantics are the same as for snprintf or strlcpy: bufSize is the size
+ * of the buffer, the string is always null-terminated, and the return value is
+ * the number of characters (not including the null terminator) that would have
+ * been written if the buffer was big enough. (So a return value >= bufSize
+ * indicates that the output was truncated)
+ *
+ * This function does not allocate memory and is async-signal-safe.
+ *
+ * Note that the underlying function for the fbstring-returning demangle is
+ * somewhat standard (abi::__cxa_demangle, which uses malloc), the underlying
+ * function for this version is less so (cplus_demangle_v3_callback from
+ * libiberty), so it is possible for the fbstring version to work, while this
+ * version returns the original, mangled name.
+ */
+size_t demangle(const char* name, char* buf, size_t bufSize);
+inline size_t demangle(const std::type_info& type, char* buf, size_t bufSize) {
+  return demangle(type.name(), buf, bufSize);
+}
+
 /**
  * Debug string for an exception: include type and what().
  */
@@ -344,6 +417,35 @@ void splitTo(const Delim& delimiter,
              OutputIterator out,
              bool ignoreEmpty = false);
 
+/*
+ * Split a string into a fixed number of pieces by delimiter. Returns 'true' if
+ * the fields were all successfully populated.
+ *
+ * Example:
+ *
+ *  folly::StringPiece name, key, value;
+ *  if (folly::split('\t', line, name, key, value))
+ *    ...
+ *
+ * The 'exact' template paremeter 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");
+ */
+template<bool exact = true,
+         class Delim,
+         class... StringPieces>
+bool split(const Delim& delimiter,
+           StringPiece input,
+           StringPiece& outHead,
+           StringPieces&... outTail);
+
 /*
  * Join list of tokens.
  *
@@ -364,6 +466,13 @@ void join(const Delim& delimiter,
   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) {
@@ -372,6 +481,14 @@ std::string join(const Delim& delimiter,
   return output;
 }
 
+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;
+}
+
 } // namespace folly
 
 // Hash functions for string and fbstring usable with e.g. hash_map