forked from OSchip/llvm-project
1858 lines
52 KiB
C++
1858 lines
52 KiB
C++
//===- unittests/AST/CommentLexer.cpp ------ Comment lexer tests ----------===//
|
|
//
|
|
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
|
|
// See https://llvm.org/LICENSE.txt for license information.
|
|
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#include "clang/AST/CommentLexer.h"
|
|
#include "clang/AST/CommentCommandTraits.h"
|
|
#include "clang/Basic/CommentOptions.h"
|
|
#include "clang/Basic/Diagnostic.h"
|
|
#include "clang/Basic/DiagnosticOptions.h"
|
|
#include "clang/Basic/FileManager.h"
|
|
#include "clang/Basic/SourceManager.h"
|
|
#include "llvm/ADT/STLExtras.h"
|
|
#include "gtest/gtest.h"
|
|
#include <vector>
|
|
|
|
using namespace llvm;
|
|
using namespace clang;
|
|
|
|
namespace clang {
|
|
namespace comments {
|
|
|
|
namespace {
|
|
class CommentLexerTest : public ::testing::Test {
|
|
protected:
|
|
CommentLexerTest()
|
|
: FileMgr(FileMgrOpts),
|
|
DiagID(new DiagnosticIDs()),
|
|
Diags(DiagID, new DiagnosticOptions, new IgnoringDiagConsumer()),
|
|
SourceMgr(Diags, FileMgr),
|
|
Traits(Allocator, CommentOptions()) {
|
|
}
|
|
|
|
FileSystemOptions FileMgrOpts;
|
|
FileManager FileMgr;
|
|
IntrusiveRefCntPtr<DiagnosticIDs> DiagID;
|
|
DiagnosticsEngine Diags;
|
|
SourceManager SourceMgr;
|
|
llvm::BumpPtrAllocator Allocator;
|
|
CommandTraits Traits;
|
|
|
|
void lexString(const char *Source, std::vector<Token> &Toks);
|
|
|
|
StringRef getCommandName(const Token &Tok) {
|
|
return Traits.getCommandInfo(Tok.getCommandID())->Name;
|
|
}
|
|
|
|
StringRef getVerbatimBlockName(const Token &Tok) {
|
|
return Traits.getCommandInfo(Tok.getVerbatimBlockID())->Name;
|
|
}
|
|
|
|
StringRef getVerbatimLineName(const Token &Tok) {
|
|
return Traits.getCommandInfo(Tok.getVerbatimLineID())->Name;
|
|
}
|
|
};
|
|
|
|
void CommentLexerTest::lexString(const char *Source,
|
|
std::vector<Token> &Toks) {
|
|
std::unique_ptr<MemoryBuffer> Buf = MemoryBuffer::getMemBuffer(Source);
|
|
FileID File = SourceMgr.createFileID(std::move(Buf));
|
|
SourceLocation Begin = SourceMgr.getLocForStartOfFile(File);
|
|
|
|
Lexer L(Allocator, Diags, Traits, Begin, Source, Source + strlen(Source));
|
|
|
|
while (1) {
|
|
Token Tok;
|
|
L.lex(Tok);
|
|
if (Tok.is(tok::eof))
|
|
break;
|
|
Toks.push_back(Tok);
|
|
}
|
|
}
|
|
|
|
} // unnamed namespace
|
|
|
|
// Empty source range should be handled.
|
|
TEST_F(CommentLexerTest, Basic1) {
|
|
const char *Source = "";
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(0U, Toks.size());
|
|
}
|
|
|
|
// Empty comments should be handled.
|
|
TEST_F(CommentLexerTest, Basic2) {
|
|
const char *Sources[] = {
|
|
"//", "///", "//!", "///<", "//!<"
|
|
};
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(1U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[0].getKind());
|
|
}
|
|
}
|
|
|
|
// Empty comments should be handled.
|
|
TEST_F(CommentLexerTest, Basic3) {
|
|
const char *Sources[] = {
|
|
"/**/", "/***/", "/*!*/", "/**<*/", "/*!<*/"
|
|
};
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(2U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[0].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[1].getKind());
|
|
}
|
|
}
|
|
|
|
// Single comment with plain text.
|
|
TEST_F(CommentLexerTest, Basic4) {
|
|
const char *Sources[] = {
|
|
"// Meow", "/// Meow", "//! Meow",
|
|
"// Meow\n", "// Meow\r\n", "//! Meow\r",
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(2U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" Meow"), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[1].getKind());
|
|
}
|
|
}
|
|
|
|
// Single comment with plain text.
|
|
TEST_F(CommentLexerTest, Basic5) {
|
|
const char *Sources[] = {
|
|
"/* Meow*/", "/** Meow*/", "/*! Meow*/"
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" Meow"), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[1].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
}
|
|
|
|
// Test newline escaping.
|
|
TEST_F(CommentLexerTest, Basic6) {
|
|
const char *Sources[] = {
|
|
"// Aaa\\\n" " Bbb\\ \n" " Ccc?" "?/\n",
|
|
"// Aaa\\\r\n" " Bbb\\ \r\n" " Ccc?" "?/\r\n",
|
|
"// Aaa\\\r" " Bbb\\ \r" " Ccc?" "?/\r"
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(10U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" Aaa"), Toks[0].getText());
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("\\"), Toks[1].getText());
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef(" Bbb"), Toks[3].getText());
|
|
ASSERT_EQ(tok::text, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef("\\"), Toks[4].getText());
|
|
ASSERT_EQ(tok::text, Toks[5].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[5].getText());
|
|
ASSERT_EQ(tok::newline, Toks[6].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[7].getKind());
|
|
ASSERT_EQ(StringRef(" Ccc?" "?/"), Toks[7].getText());
|
|
ASSERT_EQ(tok::newline, Toks[8].getKind());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[9].getKind());
|
|
}
|
|
}
|
|
|
|
// Check that we skip C-style aligned stars correctly.
|
|
TEST_F(CommentLexerTest, Basic7) {
|
|
const char *Source =
|
|
"/* Aaa\n"
|
|
" * Bbb\r\n"
|
|
"\t* Ccc\n"
|
|
" ! Ddd\n"
|
|
" * Eee\n"
|
|
" ** Fff\n"
|
|
" */";
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(15U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" Aaa"), Toks[0].getText());
|
|
ASSERT_EQ(tok::newline, Toks[1].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" Bbb"), Toks[2].getText());
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef(" Ccc"), Toks[4].getText());
|
|
ASSERT_EQ(tok::newline, Toks[5].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[6].getKind());
|
|
ASSERT_EQ(StringRef(" ! Ddd"), Toks[6].getText());
|
|
ASSERT_EQ(tok::newline, Toks[7].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[8].getKind());
|
|
ASSERT_EQ(StringRef(" Eee"), Toks[8].getText());
|
|
ASSERT_EQ(tok::newline, Toks[9].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[10].getKind());
|
|
ASSERT_EQ(StringRef("* Fff"), Toks[10].getText());
|
|
ASSERT_EQ(tok::newline, Toks[11].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[12].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[12].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[13].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[14].getKind());
|
|
}
|
|
|
|
// A command marker followed by comment end.
|
|
TEST_F(CommentLexerTest, DoxygenCommand1) {
|
|
const char *Sources[] = { "//@", "///@", "//!@" };
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(2U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef("@"), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[1].getKind());
|
|
}
|
|
}
|
|
|
|
// A command marker followed by comment end.
|
|
TEST_F(CommentLexerTest, DoxygenCommand2) {
|
|
const char *Sources[] = { "/*@*/", "/**@*/", "/*!@*/"};
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef("@"), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[1].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
}
|
|
|
|
// A command marker followed by comment end.
|
|
TEST_F(CommentLexerTest, DoxygenCommand3) {
|
|
const char *Sources[] = { "/*\\*/", "/**\\*/" };
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef("\\"), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[1].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
}
|
|
|
|
// Doxygen escape sequences.
|
|
TEST_F(CommentLexerTest, DoxygenCommand4) {
|
|
const char *Sources[] = {
|
|
"/// \\\\ \\@ \\& \\$ \\# \\< \\> \\% \\\" \\. \\::",
|
|
"/// @\\ @@ @& @$ @# @< @> @% @\" @. @::"
|
|
};
|
|
const char *Text[] = {
|
|
" ",
|
|
"\\", " ", "@", " ", "&", " ", "$", " ", "#", " ",
|
|
"<", " ", ">", " ", "%", " ", "\"", " ", ".", " ",
|
|
"::", ""
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(array_lengthof(Text), Toks.size());
|
|
|
|
for (size_t j = 0, e = Toks.size(); j != e; j++) {
|
|
if(Toks[j].is(tok::text)) {
|
|
ASSERT_EQ(StringRef(Text[j]), Toks[j].getText())
|
|
<< "index " << i;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
// A command marker followed by a non-letter that is not a part of an escape
|
|
// sequence.
|
|
TEST_F(CommentLexerTest, DoxygenCommand5) {
|
|
const char *Source = "/// \\^ \\0";
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(6U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("\\"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("^ "), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef("\\"), Toks[3].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef("0"), Toks[4].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[5].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, DoxygenCommand6) {
|
|
const char *Source = "/// \\brief Aaa.";
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::backslash_command, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("brief"), getCommandName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" Aaa."), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, DoxygenCommand7) {
|
|
const char *Source = "/// \\em\\em \\em\t\\em\n";
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(8U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::backslash_command, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("em"), getCommandName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::backslash_command, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("em"), getCommandName(Toks[2]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[3].getText());
|
|
|
|
ASSERT_EQ(tok::backslash_command, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef("em"), getCommandName(Toks[4]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[5].getKind());
|
|
ASSERT_EQ(StringRef("\t"), Toks[5].getText());
|
|
|
|
ASSERT_EQ(tok::backslash_command, Toks[6].getKind());
|
|
ASSERT_EQ(StringRef("em"), getCommandName(Toks[6]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[7].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, DoxygenCommand8) {
|
|
const char *Source = "/// @em@em @em\t@em\n";
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(8U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::at_command, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("em"), getCommandName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::at_command, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("em"), getCommandName(Toks[2]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[3].getText());
|
|
|
|
ASSERT_EQ(tok::at_command, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef("em"), getCommandName(Toks[4]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[5].getKind());
|
|
ASSERT_EQ(StringRef("\t"), Toks[5].getText());
|
|
|
|
ASSERT_EQ(tok::at_command, Toks[6].getKind());
|
|
ASSERT_EQ(StringRef("em"), getCommandName(Toks[6]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[7].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, DoxygenCommand9) {
|
|
const char *Source = "/// \\aaa\\bbb \\ccc\t\\ddd\n";
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(8U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::unknown_command, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("aaa"), Toks[1].getUnknownCommandName());
|
|
|
|
ASSERT_EQ(tok::unknown_command, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("bbb"), Toks[2].getUnknownCommandName());
|
|
|
|
ASSERT_EQ(tok::text, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[3].getText());
|
|
|
|
ASSERT_EQ(tok::unknown_command, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef("ccc"), Toks[4].getUnknownCommandName());
|
|
|
|
ASSERT_EQ(tok::text, Toks[5].getKind());
|
|
ASSERT_EQ(StringRef("\t"), Toks[5].getText());
|
|
|
|
ASSERT_EQ(tok::unknown_command, Toks[6].getKind());
|
|
ASSERT_EQ(StringRef("ddd"), Toks[6].getUnknownCommandName());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[7].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, DoxygenCommand10) {
|
|
const char *Source = "// \\c\n";
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::backslash_command, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("c"), getCommandName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, RegisterCustomBlockCommand) {
|
|
const char *Source =
|
|
"/// \\NewBlockCommand Aaa.\n"
|
|
"/// @NewBlockCommand Aaa.\n";
|
|
|
|
Traits.registerBlockCommand(StringRef("NewBlockCommand"));
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(8U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::backslash_command, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("NewBlockCommand"), getCommandName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" Aaa."), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[4].getText());
|
|
|
|
ASSERT_EQ(tok::at_command, Toks[5].getKind());
|
|
ASSERT_EQ(StringRef("NewBlockCommand"), getCommandName(Toks[5]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[6].getKind());
|
|
ASSERT_EQ(StringRef(" Aaa."), Toks[6].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[7].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, RegisterMultipleBlockCommands) {
|
|
const char *Source =
|
|
"/// \\Foo\n"
|
|
"/// \\Bar Baz\n"
|
|
"/// \\Blech quux=corge\n";
|
|
|
|
Traits.registerBlockCommand(StringRef("Foo"));
|
|
Traits.registerBlockCommand(StringRef("Bar"));
|
|
Traits.registerBlockCommand(StringRef("Blech"));
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(11U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::backslash_command, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("Foo"), getCommandName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[3].getText());
|
|
|
|
ASSERT_EQ(tok::backslash_command, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef("Bar"), getCommandName(Toks[4]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[5].getKind());
|
|
ASSERT_EQ(StringRef(" Baz"), Toks[5].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[6].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[7].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[7].getText());
|
|
|
|
ASSERT_EQ(tok::backslash_command, Toks[8].getKind());
|
|
ASSERT_EQ(StringRef("Blech"), getCommandName(Toks[8]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[9].getKind());
|
|
ASSERT_EQ(StringRef(" quux=corge"), Toks[9].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[10].getKind());
|
|
}
|
|
|
|
// Empty verbatim block.
|
|
TEST_F(CommentLexerTest, VerbatimBlock1) {
|
|
const char *Sources[] = {
|
|
"/// \\verbatim\\endverbatim\n//",
|
|
"/** \\verbatim\\endverbatim*/"
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(5U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("verbatim"), getVerbatimBlockName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::verbatim_block_end, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("endverbatim"), getVerbatimBlockName(Toks[2]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[4].getKind());
|
|
}
|
|
}
|
|
|
|
// Empty verbatim block without an end command.
|
|
TEST_F(CommentLexerTest, VerbatimBlock2) {
|
|
const char *Source = "/// \\verbatim";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("verbatim"), getVerbatimBlockName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
|
|
// Empty verbatim block without an end command.
|
|
TEST_F(CommentLexerTest, VerbatimBlock3) {
|
|
const char *Source = "/** \\verbatim*/";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("verbatim"), getVerbatimBlockName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
// Single-line verbatim block.
|
|
TEST_F(CommentLexerTest, VerbatimBlock4) {
|
|
const char *Sources[] = {
|
|
"/// Meow \\verbatim aaa \\endverbatim\n//",
|
|
"/** Meow \\verbatim aaa \\endverbatim*/"
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(6U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" Meow "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("verbatim"), getVerbatimBlockName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" aaa "), Toks[2].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_end, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef("endverbatim"), getVerbatimBlockName(Toks[3]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[4].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[5].getKind());
|
|
}
|
|
}
|
|
|
|
// Single-line verbatim block without an end command.
|
|
TEST_F(CommentLexerTest, VerbatimBlock5) {
|
|
const char *Sources[] = {
|
|
"/// Meow \\verbatim aaa \n//",
|
|
"/** Meow \\verbatim aaa */"
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(5U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" Meow "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("verbatim"), getVerbatimBlockName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" aaa "), Toks[2].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[4].getKind());
|
|
}
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, VerbatimBlock6) {
|
|
const char *Source =
|
|
"// \\verbatim\n"
|
|
"// Aaa\n"
|
|
"//\n"
|
|
"// Bbb\n"
|
|
"// \\endverbatim\n";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(10U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("verbatim"), getVerbatimBlockName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef(" Aaa"), Toks[3].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[4].getKind());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[5].getKind());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[6].getKind());
|
|
ASSERT_EQ(StringRef(" Bbb"), Toks[6].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[7].getKind());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_end, Toks[8].getKind());
|
|
ASSERT_EQ(StringRef("endverbatim"), getVerbatimBlockName(Toks[8]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[9].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, VerbatimBlock7) {
|
|
const char *Source =
|
|
"/* \\verbatim\n"
|
|
" * Aaa\n"
|
|
" *\n"
|
|
" * Bbb\n"
|
|
" * \\endverbatim\n"
|
|
" */";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(10U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("verbatim"), getVerbatimBlockName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" Aaa"), Toks[2].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef(""), Toks[3].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef(" Bbb"), Toks[4].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_end, Toks[5].getKind());
|
|
ASSERT_EQ(StringRef("endverbatim"), getVerbatimBlockName(Toks[5]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[6].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[7].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[7].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[8].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[9].getKind());
|
|
}
|
|
|
|
// Complex test for verbatim blocks.
|
|
TEST_F(CommentLexerTest, VerbatimBlock8) {
|
|
const char *Source =
|
|
"/* Meow \\verbatim aaa\\$\\@\n"
|
|
"bbb \\endverbati\r"
|
|
"ccc\r\n"
|
|
"ddd \\endverbatim Blah \\verbatim eee\n"
|
|
"\\endverbatim BlahBlah*/";
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(14U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" Meow "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("verbatim"), getVerbatimBlockName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" aaa\\$\\@"), Toks[2].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef("bbb \\endverbati"), Toks[3].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef("ccc"), Toks[4].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[5].getKind());
|
|
ASSERT_EQ(StringRef("ddd "), Toks[5].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_end, Toks[6].getKind());
|
|
ASSERT_EQ(StringRef("endverbatim"), getVerbatimBlockName(Toks[6]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[7].getKind());
|
|
ASSERT_EQ(StringRef(" Blah "), Toks[7].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[8].getKind());
|
|
ASSERT_EQ(StringRef("verbatim"), getVerbatimBlockName(Toks[8]));
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[9].getKind());
|
|
ASSERT_EQ(StringRef(" eee"), Toks[9].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_end, Toks[10].getKind());
|
|
ASSERT_EQ(StringRef("endverbatim"), getVerbatimBlockName(Toks[10]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[11].getKind());
|
|
ASSERT_EQ(StringRef(" BlahBlah"), Toks[11].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[12].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[13].getKind());
|
|
}
|
|
|
|
// LaTeX verbatim blocks.
|
|
TEST_F(CommentLexerTest, VerbatimBlock9) {
|
|
const char *Source =
|
|
"/// \\f$ Aaa \\f$ \\f[ Bbb \\f] \\f{ Ccc \\f} \\f( Ddd \\f)";
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(17U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("f$"), getVerbatimBlockName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" Aaa "), Toks[2].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_end, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef("f$"), getVerbatimBlockName(Toks[3]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[4].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[5].getKind());
|
|
ASSERT_EQ(StringRef("f["), getVerbatimBlockName(Toks[5]));
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[6].getKind());
|
|
ASSERT_EQ(StringRef(" Bbb "), Toks[6].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_end, Toks[7].getKind());
|
|
ASSERT_EQ(StringRef("f]"), getVerbatimBlockName(Toks[7]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[8].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[8].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[9].getKind());
|
|
ASSERT_EQ(StringRef("f{"), getVerbatimBlockName(Toks[9]));
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[10].getKind());
|
|
ASSERT_EQ(StringRef(" Ccc "), Toks[10].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_end, Toks[11].getKind());
|
|
ASSERT_EQ(StringRef("f}"), getVerbatimBlockName(Toks[11]));
|
|
|
|
ASSERT_EQ(tok::text, Toks[12].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[12].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_begin, Toks[13].getKind());
|
|
ASSERT_EQ(StringRef("f("), getVerbatimBlockName(Toks[13]));
|
|
|
|
ASSERT_EQ(tok::verbatim_block_line, Toks[14].getKind());
|
|
ASSERT_EQ(StringRef(" Ddd "), Toks[14].getVerbatimBlockText());
|
|
|
|
ASSERT_EQ(tok::verbatim_block_end, Toks[15].getKind());
|
|
ASSERT_EQ(StringRef("f)"), getVerbatimBlockName(Toks[15]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[16].getKind());
|
|
}
|
|
|
|
// Empty verbatim line.
|
|
TEST_F(CommentLexerTest, VerbatimLine1) {
|
|
const char *Sources[] = {
|
|
"/// \\fn\n//",
|
|
"/** \\fn*/"
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_line_name, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("fn"), getVerbatimLineName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
}
|
|
|
|
// Verbatim line with Doxygen escape sequences, which should not be expanded.
|
|
TEST_F(CommentLexerTest, VerbatimLine2) {
|
|
const char *Sources[] = {
|
|
"/// \\fn void *foo(const char *zzz = \"\\$\");\n//",
|
|
"/** \\fn void *foo(const char *zzz = \"\\$\");*/"
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(5U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_line_name, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("fn"), getVerbatimLineName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::verbatim_line_text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" void *foo(const char *zzz = \"\\$\");"),
|
|
Toks[2].getVerbatimLineText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[4].getKind());
|
|
}
|
|
}
|
|
|
|
// Verbatim line should not eat anything from next source line.
|
|
TEST_F(CommentLexerTest, VerbatimLine3) {
|
|
const char *Source =
|
|
"/** \\fn void *foo(const char *zzz = \"\\$\");\n"
|
|
" * Meow\n"
|
|
" */";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(9U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::verbatim_line_name, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("fn"), getVerbatimLineName(Toks[1]));
|
|
|
|
ASSERT_EQ(tok::verbatim_line_text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" void *foo(const char *zzz = \"\\$\");"),
|
|
Toks[2].getVerbatimLineText());
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef(" Meow"), Toks[4].getText());
|
|
ASSERT_EQ(tok::newline, Toks[5].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[6].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[6].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[7].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[8].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML1) {
|
|
const char *Source =
|
|
"// <";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("<"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML2) {
|
|
const char *Source =
|
|
"// a<2";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" a"), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("<"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("2"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML3) {
|
|
const char *Source =
|
|
"// < img";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("<"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" img"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML4) {
|
|
const char *Sources[] = {
|
|
"// <img",
|
|
"// <img "
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML5) {
|
|
const char *Source =
|
|
"// <img 42";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("42"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML6) {
|
|
const char *Source = "// <img> Meow";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(5U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::html_greater, Toks[2].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef(" Meow"), Toks[3].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[4].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML7) {
|
|
const char *Source = "// <img=";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("="), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML8) {
|
|
const char *Source = "// <img src=> Meow";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(7U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::html_ident, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("src"), Toks[2].getHTMLIdent());
|
|
|
|
ASSERT_EQ(tok::html_equals, Toks[3].getKind());
|
|
|
|
ASSERT_EQ(tok::html_greater, Toks[4].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[5].getKind());
|
|
ASSERT_EQ(StringRef(" Meow"), Toks[5].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[6].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML9) {
|
|
const char *Sources[] = {
|
|
"// <img src",
|
|
"// <img src "
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::html_ident, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("src"), Toks[2].getHTMLIdent());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML10) {
|
|
const char *Sources[] = {
|
|
"// <img src=",
|
|
"// <img src ="
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(5U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::html_ident, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("src"), Toks[2].getHTMLIdent());
|
|
|
|
ASSERT_EQ(tok::html_equals, Toks[3].getKind());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[4].getKind());
|
|
}
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML11) {
|
|
const char *Sources[] = {
|
|
"// <img src=\"",
|
|
"// <img src = \"",
|
|
"// <img src=\'",
|
|
"// <img src = \'"
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(6U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::html_ident, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("src"), Toks[2].getHTMLIdent());
|
|
|
|
ASSERT_EQ(tok::html_equals, Toks[3].getKind());
|
|
|
|
ASSERT_EQ(tok::html_quoted_string, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef(""), Toks[4].getHTMLQuotedString());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[5].getKind());
|
|
}
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML12) {
|
|
const char *Source = "// <img src=@";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(6U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::html_ident, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("src"), Toks[2].getHTMLIdent());
|
|
|
|
ASSERT_EQ(tok::html_equals, Toks[3].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef("@"), Toks[4].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[5].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML13) {
|
|
const char *Sources[] = {
|
|
"// <img src=\"val\\\"\\'val",
|
|
"// <img src=\"val\\\"\\'val\"",
|
|
"// <img src=\'val\\\"\\'val",
|
|
"// <img src=\'val\\\"\\'val\'"
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(6U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::html_ident, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("src"), Toks[2].getHTMLIdent());
|
|
|
|
ASSERT_EQ(tok::html_equals, Toks[3].getKind());
|
|
|
|
ASSERT_EQ(tok::html_quoted_string, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef("val\\\"\\'val"), Toks[4].getHTMLQuotedString());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[5].getKind());
|
|
}
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML14) {
|
|
const char *Sources[] = {
|
|
"// <img src=\"val\\\"\\'val\">",
|
|
"// <img src=\'val\\\"\\'val\'>"
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(7U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::html_ident, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("src"), Toks[2].getHTMLIdent());
|
|
|
|
ASSERT_EQ(tok::html_equals, Toks[3].getKind());
|
|
|
|
ASSERT_EQ(tok::html_quoted_string, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef("val\\\"\\'val"), Toks[4].getHTMLQuotedString());
|
|
|
|
ASSERT_EQ(tok::html_greater, Toks[5].getKind());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[6].getKind());
|
|
}
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML15) {
|
|
const char *Sources[] = {
|
|
"// <img/>",
|
|
"// <img />"
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::html_slash_greater, Toks[2].getKind());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML16) {
|
|
const char *Sources[] = {
|
|
"// <img/ Aaa",
|
|
"// <img / Aaa"
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(5U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_start_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagStartName());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("/"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[3].getKind());
|
|
ASSERT_EQ(StringRef(" Aaa"), Toks[3].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[4].getKind());
|
|
}
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML17) {
|
|
const char *Source = "// </";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("</"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML18) {
|
|
const char *Source = "// </@";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("</"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("@"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTML19) {
|
|
const char *Source = "// </img";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::html_end_tag, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("img"), Toks[1].getHTMLTagEndName());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, NotAKnownHTMLTag1) {
|
|
const char *Source = "// <tag>";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("<tag"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(">"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, NotAKnownHTMLTag2) {
|
|
const char *Source = "// </tag>";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("</tag"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(">"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences1) {
|
|
const char *Source = "// &";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("&"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences2) {
|
|
const char *Source = "// &!";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("&"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("!"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences3) {
|
|
const char *Source = "// &";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("&"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences4) {
|
|
const char *Source = "// &!";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("&"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("!"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences5) {
|
|
const char *Source = "// &#";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("&#"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences6) {
|
|
const char *Source = "// &#a";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("&#"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("a"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences7) {
|
|
const char *Source = "// *";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("*"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences8) {
|
|
const char *Source = "// *a";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("*"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("a"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences9) {
|
|
const char *Source = "// &#x";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("&#x"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences10) {
|
|
const char *Source = "// &#xz";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("&#x"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("z"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences11) {
|
|
const char *Source = "// «";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("«"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences12) {
|
|
const char *Source = "// «z";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("«"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("z"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences13) {
|
|
const char *Source = "// &";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("&"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences14) {
|
|
const char *Source = "// &<";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("&"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef("<"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences15) {
|
|
const char *Source = "// & meow";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(4U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("&"), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" meow"), Toks[2].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, HTMLCharacterReferences16) {
|
|
const char *Sources[] = {
|
|
"// =",
|
|
"// =",
|
|
"// =",
|
|
"// ="
|
|
};
|
|
|
|
for (size_t i = 0, e = array_lengthof(Sources); i != e; i++) {
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Sources[i], Toks);
|
|
|
|
ASSERT_EQ(3U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" "), Toks[0].getText());
|
|
|
|
ASSERT_EQ(tok::text, Toks[1].getKind());
|
|
ASSERT_EQ(StringRef("="), Toks[1].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[2].getKind());
|
|
}
|
|
}
|
|
|
|
TEST_F(CommentLexerTest, MultipleComments) {
|
|
const char *Source =
|
|
"// Aaa\n"
|
|
"/// Bbb\n"
|
|
"/* Ccc\n"
|
|
" * Ddd*/\n"
|
|
"/** Eee*/";
|
|
|
|
std::vector<Token> Toks;
|
|
|
|
lexString(Source, Toks);
|
|
|
|
ASSERT_EQ(12U, Toks.size());
|
|
|
|
ASSERT_EQ(tok::text, Toks[0].getKind());
|
|
ASSERT_EQ(StringRef(" Aaa"), Toks[0].getText());
|
|
ASSERT_EQ(tok::newline, Toks[1].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[2].getKind());
|
|
ASSERT_EQ(StringRef(" Bbb"), Toks[2].getText());
|
|
ASSERT_EQ(tok::newline, Toks[3].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[4].getKind());
|
|
ASSERT_EQ(StringRef(" Ccc"), Toks[4].getText());
|
|
ASSERT_EQ(tok::newline, Toks[5].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[6].getKind());
|
|
ASSERT_EQ(StringRef(" Ddd"), Toks[6].getText());
|
|
ASSERT_EQ(tok::newline, Toks[7].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[8].getKind());
|
|
|
|
ASSERT_EQ(tok::text, Toks[9].getKind());
|
|
ASSERT_EQ(StringRef(" Eee"), Toks[9].getText());
|
|
|
|
ASSERT_EQ(tok::newline, Toks[10].getKind());
|
|
ASSERT_EQ(tok::newline, Toks[11].getKind());
|
|
}
|
|
|
|
} // end namespace comments
|
|
} // end namespace clang
|
|
|