| /* -*- Mode:C++; c-file-style:"gnu"; indent-tabs-mode:nil; -*- */ |
| /** |
| * Copyright (c) 2013-2016 Regents of the University of California. |
| * |
| * This file is part of ndn-cxx library (NDN C++ library with eXperimental eXtensions). |
| * |
| * ndn-cxx library is free software: you can redistribute it and/or modify it under the |
| * terms of the GNU Lesser General Public License as published by the Free Software |
| * Foundation, either version 3 of the License, or (at your option) any later version. |
| * |
| * ndn-cxx library is distributed in the hope that it will be useful, but WITHOUT ANY |
| * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A |
| * PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details. |
| * |
| * You should have received copies of the GNU General Public License and GNU Lesser |
| * General Public License along with ndn-cxx, e.g., in COPYING.md file. If not, see |
| * <http://www.gnu.org/licenses/>. |
| * |
| * See AUTHORS.md for complete list of ndn-cxx authors and contributors. |
| * |
| * @author Jeff Thompson <jefft0@remap.ucla.edu> |
| * @author Alexander Afanasyev <http://lasr.cs.ucla.edu/afanasyev/index.html> |
| * @author Zhenkai Zhu <http://irl.cs.ucla.edu/~zhenkai/> |
| */ |
| |
| #include "name.hpp" |
| |
| #include "util/time.hpp" |
| #include "util/string-helper.hpp" |
| #include "encoding/block.hpp" |
| #include "encoding/encoding-buffer.hpp" |
| |
| #include <boost/functional/hash.hpp> |
| |
| namespace ndn { |
| |
| BOOST_CONCEPT_ASSERT((boost::EqualityComparable<Name>)); |
| BOOST_CONCEPT_ASSERT((WireEncodable<Name>)); |
| BOOST_CONCEPT_ASSERT((WireEncodableWithEncodingBuffer<Name>)); |
| BOOST_CONCEPT_ASSERT((WireDecodable<Name>)); |
| static_assert(std::is_base_of<tlv::Error, Name::Error>::value, |
| "Name::Error must inherit from tlv::Error"); |
| |
| const size_t Name::npos = std::numeric_limits<size_t>::max(); |
| |
| Name::Name() |
| : m_nameBlock(tlv::Name) |
| { |
| } |
| |
| Name::Name(const Block& wire) |
| { |
| m_nameBlock = wire; |
| m_nameBlock.parse(); |
| } |
| |
| Name::Name(const char* uri) |
| { |
| construct(uri); |
| } |
| |
| Name::Name(const std::string& uri) |
| { |
| construct(uri.c_str()); |
| } |
| |
| Name |
| Name::deepCopy() const |
| { |
| Name copiedName(*this); |
| copiedName.m_nameBlock.resetWire(); |
| copiedName.wireEncode(); // "compress" the underlying buffer |
| return copiedName; |
| } |
| |
| template<encoding::Tag TAG> |
| size_t |
| Name::wireEncode(EncodingImpl<TAG>& encoder) const |
| { |
| size_t totalLength = 0; |
| |
| for (const_reverse_iterator i = rbegin(); i != rend(); ++i) |
| { |
| totalLength += i->wireEncode(encoder); |
| } |
| |
| totalLength += encoder.prependVarNumber(totalLength); |
| totalLength += encoder.prependVarNumber(tlv::Name); |
| return totalLength; |
| } |
| |
| template size_t |
| Name::wireEncode<encoding::EncoderTag>(EncodingImpl<encoding::EncoderTag>& encoder) const; |
| |
| template size_t |
| Name::wireEncode<encoding::EstimatorTag>(EncodingImpl<encoding::EstimatorTag>& encoder) const; |
| |
| const Block& |
| Name::wireEncode() const |
| { |
| if (m_nameBlock.hasWire()) |
| return m_nameBlock; |
| |
| EncodingEstimator estimator; |
| size_t estimatedSize = wireEncode(estimator); |
| |
| EncodingBuffer buffer(estimatedSize, 0); |
| wireEncode(buffer); |
| |
| m_nameBlock = buffer.block(); |
| m_nameBlock.parse(); |
| |
| return m_nameBlock; |
| } |
| |
| void |
| Name::wireDecode(const Block& wire) |
| { |
| if (wire.type() != tlv::Name) |
| BOOST_THROW_EXCEPTION(tlv::Error("Unexpected TLV type when decoding Name")); |
| |
| m_nameBlock = wire; |
| m_nameBlock.parse(); |
| } |
| |
| void |
| Name::construct(const char* uriOrig) |
| { |
| clear(); |
| |
| std::string uri = uriOrig; |
| trim(uri); |
| if (uri.size() == 0) |
| return; |
| |
| size_t iColon = uri.find(':'); |
| if (iColon != std::string::npos) { |
| // Make sure the colon came before a '/'. |
| size_t iFirstSlash = uri.find('/'); |
| if (iFirstSlash == std::string::npos || iColon < iFirstSlash) { |
| // Omit the leading protocol such as ndn: |
| uri.erase(0, iColon + 1); |
| trim(uri); |
| } |
| } |
| |
| // Trim the leading slash and possibly the authority. |
| if (uri[0] == '/') { |
| if (uri.size() >= 2 && uri[1] == '/') { |
| // Strip the authority following "//". |
| size_t iAfterAuthority = uri.find('/', 2); |
| if (iAfterAuthority == std::string::npos) |
| // Unusual case: there was only an authority. |
| return; |
| else { |
| uri.erase(0, iAfterAuthority + 1); |
| trim(uri); |
| } |
| } |
| else { |
| uri.erase(0, 1); |
| trim(uri); |
| } |
| } |
| |
| size_t iComponentStart = 0; |
| |
| // Unescape the components. |
| while (iComponentStart < uri.size()) { |
| size_t iComponentEnd = uri.find("/", iComponentStart); |
| if (iComponentEnd == std::string::npos) |
| iComponentEnd = uri.size(); |
| |
| append(Component::fromEscapedString(&uri[0], iComponentStart, iComponentEnd)); |
| iComponentStart = iComponentEnd + 1; |
| } |
| } |
| |
| void |
| Name::set(const char* uri) |
| { |
| *this = Name(uri); |
| } |
| |
| void |
| Name::set(const std::string& uri) |
| { |
| *this = Name(uri); |
| } |
| |
| std::string |
| Name::toUri() const |
| { |
| std::ostringstream os; |
| os << *this; |
| return os.str(); |
| } |
| |
| Name& |
| Name::append(const PartialName& name) |
| { |
| if (&name == this) |
| // Copying from this name, so need to make a copy first. |
| return append(PartialName(name)); |
| |
| for (size_t i = 0; i < name.size(); ++i) |
| append(name.at(i)); |
| |
| return *this; |
| } |
| |
| Name& |
| Name::appendNumber(uint64_t number) |
| { |
| m_nameBlock.push_back(Component::fromNumber(number)); |
| return *this; |
| } |
| |
| Name& |
| Name::appendNumberWithMarker(uint8_t marker, uint64_t number) |
| { |
| m_nameBlock.push_back(Component::fromNumberWithMarker(marker, number)); |
| return *this; |
| } |
| |
| Name& |
| Name::appendVersion(uint64_t version) |
| { |
| m_nameBlock.push_back(Component::fromVersion(version)); |
| return *this; |
| } |
| |
| Name& |
| Name::appendVersion() |
| { |
| appendVersion(time::toUnixTimestamp(time::system_clock::now()).count()); |
| return *this; |
| } |
| |
| Name& |
| Name::appendSegment(uint64_t segmentNo) |
| { |
| m_nameBlock.push_back(Component::fromSegment(segmentNo)); |
| return *this; |
| } |
| |
| Name& |
| Name::appendSegmentOffset(uint64_t offset) |
| { |
| m_nameBlock.push_back(Component::fromSegmentOffset(offset)); |
| return *this; |
| } |
| |
| Name& |
| Name::appendTimestamp(const time::system_clock::TimePoint& timePoint) |
| { |
| m_nameBlock.push_back(Component::fromTimestamp(timePoint)); |
| return *this; |
| } |
| |
| Name& |
| Name::appendSequenceNumber(uint64_t seqNo) |
| { |
| m_nameBlock.push_back(Component::fromSequenceNumber(seqNo)); |
| return *this; |
| } |
| |
| Name& |
| Name::appendImplicitSha256Digest(const ConstBufferPtr& digest) |
| { |
| m_nameBlock.push_back(Component::fromImplicitSha256Digest(digest)); |
| return *this; |
| } |
| |
| Name& |
| Name::appendImplicitSha256Digest(const uint8_t* digest, size_t digestSize) |
| { |
| m_nameBlock.push_back(Component::fromImplicitSha256Digest(digest, digestSize)); |
| return *this; |
| } |
| |
| PartialName |
| Name::getSubName(ssize_t iStartComponent, size_t nComponents) const |
| { |
| PartialName result; |
| |
| ssize_t iStart = iStartComponent < 0 ? this->size() + iStartComponent : iStartComponent; |
| size_t iEnd = this->size(); |
| |
| iStart = std::max(iStart, static_cast<ssize_t>(0)); |
| |
| if (nComponents != npos) |
| iEnd = std::min(this->size(), iStart + nComponents); |
| |
| for (size_t i = iStart; i < iEnd; ++i) |
| result.append(at(i)); |
| |
| return result; |
| } |
| |
| Name |
| Name::getSuccessor() const |
| { |
| if (empty()) { |
| static uint8_t firstValue[] = { 0 }; |
| Name firstName; |
| firstName.append(firstValue, 1); |
| return firstName; |
| } |
| |
| return getPrefix(-1).append(get(-1).getSuccessor()); |
| } |
| |
| bool |
| Name::equals(const Name& name) const |
| { |
| if (size() != name.size()) |
| return false; |
| |
| for (size_t i = 0; i < size(); ++i) { |
| if (at(i) != name.at(i)) |
| return false; |
| } |
| |
| return true; |
| } |
| |
| bool |
| Name::isPrefixOf(const Name& name) const |
| { |
| // This name is longer than the name we are checking against. |
| if (size() > name.size()) |
| return false; |
| |
| // Check if at least one of given components doesn't match. |
| for (size_t i = 0; i < size(); ++i) { |
| if (at(i) != name.at(i)) |
| return false; |
| } |
| |
| return true; |
| } |
| |
| int |
| Name::compare(size_t pos1, size_t count1, const Name& other, size_t pos2, size_t count2) const |
| { |
| count1 = std::min(count1, this->size() - pos1); |
| count2 = std::min(count2, other.size() - pos2); |
| size_t count = std::min(count1, count2); |
| |
| for (size_t i = 0; i < count; ++i) { |
| int comp = this->at(pos1 + i).compare(other.at(pos2 + i)); |
| if (comp != 0) { // i-th component differs |
| return comp; |
| } |
| } |
| // [pos1, pos1+count) of this Name equals [pos2, pos2+count) of other Name |
| return count1 - count2; |
| } |
| |
| std::ostream& |
| operator<<(std::ostream& os, const Name& name) |
| { |
| if (name.empty()) |
| { |
| os << "/"; |
| } |
| else |
| { |
| for (Name::const_iterator i = name.begin(); i != name.end(); i++) { |
| os << "/"; |
| i->toUri(os); |
| } |
| } |
| return os; |
| } |
| |
| std::istream& |
| operator>>(std::istream& is, Name& name) |
| { |
| std::string inputString; |
| is >> inputString; |
| name = Name(inputString); |
| |
| return is; |
| } |
| |
| } // namespace ndn |
| |
| namespace std { |
| size_t |
| hash<ndn::Name>::operator()(const ndn::Name& name) const |
| { |
| return boost::hash_range(name.wireEncode().wire(), |
| name.wireEncode().wire() + name.wireEncode().size()); |
| } |
| |
| } // namespace std |