Revert "Revert "URI parsing in folly""

Summary:
Now that the third_party link was updated in
https://phabricator.fb.com/D778617, we're good.

Test Plan: fbconfig -r thrift && fbmake runtests_opt

Reviewed By: chip@fb.com

FB internal diff: D778707

folly/String-inl.h

@@ -149,6 +149,95 @@ void cUnescape(StringPiece str, String& out, bool strict) {
   out.append(&*last, p - last);
 }
 
+namespace detail {
+// Map from character code to escape mode:
+// 0 = pass through
+// 1 = unused
+// 2 = pass through in PATH mode
+// 3 = space, replace with '+' in QUERY mode
+// 4 = percent-encode
+extern const unsigned char uriEscapeTable[];
+}  // namespace detail
+
+template <class String>
+void uriEscape(StringPiece str, String& out, UriEscapeMode mode) {
+  static const char hexValues[] = "0123456789abcdef";
+  char esc[3];
+  esc[0] = '%';
+  // Preallocate assuming that 25% of the input string will be escaped
+  out.reserve(out.size() + str.size() + 3 * (str.size() / 4));
+  auto p = str.begin();
+  auto last = p;  // last regular character
+  // We advance over runs of passthrough characters and copy them in one go;
+  // this is faster than calling push_back repeatedly.
+  unsigned char minEncode = static_cast<unsigned char>(mode);
+  while (p != str.end()) {
+    char c = *p;
+    unsigned char v = static_cast<unsigned char>(c);
+    unsigned char discriminator = detail::uriEscapeTable[v];
+    if (LIKELY(discriminator <= minEncode)) {
+      ++p;
+    } else if (mode == UriEscapeMode::QUERY && discriminator == 3) {
+      out.append(&*last, p - last);
+      out.push_back('+');
+      ++p;
+      last = p;
+    } else {
+      out.append(&*last, p - last);
+      esc[1] = hexValues[v >> 4];
+      esc[2] = hexValues[v & 0x0f];
+      out.append(esc, 3);
+      ++p;
+      last = p;
+    }
+  }
+  out.append(&*last, p - last);
+}
+
+template <class String>
+void uriUnescape(StringPiece str, String& out, UriEscapeMode mode) {
+  out.reserve(out.size() + str.size());
+  auto p = str.begin();
+  auto last = p;
+  // We advance over runs of passthrough characters and copy them in one go;
+  // this is faster than calling push_back repeatedly.
+  while (p != str.end()) {
+    char c = *p;
+    unsigned char v = static_cast<unsigned char>(v);
+    switch (c) {
+    case '%':
+      {
+        if (UNLIKELY(std::distance(p, str.end()) < 3)) {
+          throw std::invalid_argument("incomplete percent encode sequence");
+        }
+        auto h1 = detail::hexTable[static_cast<unsigned char>(p[1])];
+        auto h2 = detail::hexTable[static_cast<unsigned char>(p[2])];
+        if (UNLIKELY(h1 == 16 || h2 == 16)) {
+          throw std::invalid_argument("invalid percent encode sequence");
+        }
+        out.append(&*last, p - last);
+        out.push_back((h1 << 4) | h2);
+        p += 3;
+        last = p;
+        break;
+      }
+    case '+':
+      if (mode == UriEscapeMode::QUERY) {
+        out.append(&*last, p - last);
+        out.push_back(' ');
+        ++p;
+        last = p;
+        break;
+      }
+      // else fallthrough
+    default:
+      ++p;
+      break;
+    }
+  }
+  out.append(&*last, p - last);
+}
+
 namespace detail {
 
 /*

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

folly/Uri-inl.h

@@ -0,0 +1,49 @@
+/*
+ * Copyright 2013 Facebook, Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef FOLLY_URI_H_
+#error This file may only be included from folly/Uri.h
+#endif
+
+#include "folly/Conv.h"
+
+namespace folly {
+
+template <class String>
+String Uri::toString() const {
+  String str;
+  toAppend(scheme_, "://", &str);
+  if (!password_.empty()) {
+    toAppend(username_, ":", password_, "@", &str);
+  } else if (!username_.empty()) {
+    toAppend(username_, "@", &str);
+  }
+  toAppend(host_, &str);
+  if (port_ != 0) {
+    toAppend(":", port_, &str);
+  }
+  toAppend(path_, &str);
+  if (!query_.empty()) {
+    toAppend("?", query_, &str);
+  }
+  if (!fragment_.empty()) {
+    toAppend("#", fragment_, &str);
+  }
+  return str;
+}
+
+}  // namespace folly
+

folly/Uri.cpp

@@ -0,0 +1,95 @@
+/*
+ * Copyright 2013 Facebook, Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include "folly/Uri.h"
+
+#include <ctype.h>
+#include <boost/regex.hpp>
+
+namespace folly {
+
+namespace {
+
+fbstring submatch(const boost::cmatch& m, size_t idx) {
+  auto& sub = m[idx];
+  return fbstring(sub.first, sub.second);
+}
+
+template <class String>
+void toLower(String& s) {
+  for (auto& c : s) {
+    c = tolower(c);
+  }
+}
+
+}  // namespace
+
+Uri::Uri(StringPiece str) : port_(0) {
+  static const boost::regex uriRegex(
+      "([a-zA-Z][a-zA-Z0-9+.-]*):"  // scheme:
+      "([^?#]*)"                    // authority and path
+      "(?:\\?([^#]*))?"             // ?query
+      "(?:#(.*))?");                // #fragment
+  static const boost::regex authorityAndPathRegex("//([^/]*)(/.*)?");
+
+  boost::cmatch match;
+  if (UNLIKELY(!boost::regex_match(str.begin(), str.end(), match, uriRegex))) {
+    throw std::invalid_argument("invalid URI");
+  }
+
+  scheme_ = submatch(match, 1);
+  toLower(scheme_);
+
+  StringPiece authorityAndPath(match[2].first, match[2].second);
+  boost::cmatch authorityAndPathMatch;
+  if (!boost::regex_match(authorityAndPath.begin(),
+                          authorityAndPath.end(),
+                          authorityAndPathMatch,
+                          authorityAndPathRegex)) {
+    // Does not start with //, doesn't have authority
+    path_ = authorityAndPath.fbstr();
+  } else {
+    static const boost::regex authorityRegex(
+        "(?:([^@:]*)(?::([^@]*))?@)?"  // username, password
+        "(\\[[^\\]]*\\]|[^\\[:]*)"     // host (IP-literal, dotted-IPv4, or
+                                       // named host)
+        "(?::(\\d*))?");               // port
+
+    auto authority = authorityAndPathMatch[1];
+    boost::cmatch authorityMatch;
+    if (!boost::regex_match(authority.first,
+                            authority.second,
+                            authorityMatch,
+                            authorityRegex)) {
+      throw std::invalid_argument("invalid URI authority");
+    }
+
+    StringPiece port(authorityMatch[4].first, authorityMatch[4].second);
+    if (!port.empty()) {
+      port_ = to<uint32_t>(port);
+    }
+
+    username_ = submatch(authorityMatch, 1);
+    password_ = submatch(authorityMatch, 2);
+    host_ = submatch(authorityMatch, 3);
+    path_ = submatch(authorityAndPathMatch, 2);
+  }
+
+  query_ = submatch(match, 3);
+  fragment_ = submatch(match, 4);
+}
+
+}  // namespace folly

folly/Uri.h

@@ -0,0 +1,77 @@
+/*
+ * Copyright 2013 Facebook, Inc.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *   http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef FOLLY_URI_H_
+#define FOLLY_URI_H_
+
+#include "folly/String.h"
+
+namespace folly {
+
+/**
+ * Class representing a URI.
+ *
+ * Consider http://www.facebook.com/foo/bar?key=foo#anchor
+ *
+ * The URI is broken down into its parts: scheme ("http"), authority
+ * (ie. host and port, in most cases: "www.facebook.com"), path
+ * ("/foo/bar"), query ("key=foo") and fragment ("anchor").  The scheme is
+ * lower-cased.
+ *
+ * If this Uri represents a URL, note that, to prevent ambiguity, the component
+ * parts are NOT percent-decoded; you should do this yourself with
+ * uriUnescape() (for the authority and path) and uriUnescape(...,
+ * UriEscapeMode::QUERY) (for the query, but probably only after splitting at
+ * '&' to identify the individual parameters).
+ */
+class Uri {
+ public:
+  /**
+   * Parse a Uri from a string.  Throws std::invalid_argument on parse error.
+   */
+  explicit Uri(StringPiece str);
+
+  const fbstring& scheme() const { return scheme_; }
+  const fbstring& username() const { return username_; }
+  const fbstring& password() const { return password_; }
+  const fbstring& host() const { return host_; }
+  uint32_t port() const { return port_; }
+  const fbstring& path() const { return path_; }
+  const fbstring& query() const { return query_; }
+  const fbstring& fragment() const { return fragment_; }
+
+  template <class String>
+  String toString() const;
+
+  std::string str() const { return toString<std::string>(); }
+  fbstring fbstr() const { return toString<fbstring>(); }
+
+ private:
+  fbstring scheme_;
+  fbstring username_;
+  fbstring password_;
+  fbstring host_;
+  uint32_t port_;
+  fbstring path_;
+  fbstring query_;
+  fbstring fragment_;
+};
+
+}  // namespace folly
+
+#include "folly/Uri-inl.h"
+
+#endif /* FOLLY_URI_H_ */