#ifndef PARSER_H
#define PARSER_H
-#include <vector>
#include <string>
#include <utility>
+#include <vector>
+#include "support/docstream.h"
namespace lyx {
catInvalid // 15 <delete>
};
-
-CatCode catcode(unsigned char c);
+enum cat_type {
+ NORMAL_CATCODES,
+ VERBATIM_CATCODES,
+ UNDECIDED_CATCODES
+};
enum {
class Token {
public:
///
- Token() : cs_(), char_(0), cat_(catIgnore) {}
+ Token() : cs_(), cat_(catIgnore) {}
///
- Token(char c, CatCode cat) : cs_(), char_(c), cat_(cat) {}
- ///
- Token(std::string const & cs, CatCode cat) : cs_(cs), char_(0), cat_(cat) {}
+ Token(docstring const & cs, CatCode cat) : cs_(to_utf8(cs)), cat_(cat) {}
- ///
+ /// Returns the token as string
std::string const & cs() const { return cs_; }
/// Returns the catcode of the token
CatCode cat() const { return cat_; }
- ///
- char character() const { return char_; }
- /// Returns the token as string
- std::string asString() const;
+ /** Get the character of tokens that were constructed from a single
+ * character input or a two character input and cat_ == catEscape.
+ * FIXME: The intended usage is not clear. The Token class in
+ * ../mathed/MathParser.cpp (which is the anchestor of this
+ * class) uses a separate char member for this method. I
+ * believe that the intended usage is to not cover tokens with
+ * catEscape or catComment, e.g. \code
+ * return (cs_.empty() || cat_ == catEscape || cat_ == catComment) ? 0 : cs_[0];
+ * \endcode
+ * All usages of this method should be checked. gb 2011-01-05
+ */
+ char character() const { return cs_.empty() ? 0 : cs_[0]; }
/// Returns the token verbatim
std::string asInput() const;
+ /// Is the token an alphanumerical character?
+ bool isAlnumASCII() const;
private:
///
std::string cs_;
///
- char char_;
- ///
CatCode cat_;
};
std::ostream & operator<<(std::ostream & os, Token const & t);
+#ifdef FILEDEBUG
+extern void debugToken(std::ostream & os, Token const & t, unsigned int flags);
+#endif
+
+/// A docstream version that supports putback even when not buffered
+class iparserdocstream
+{
+public:
+ typedef idocstream::int_type int_type;
+
+ iparserdocstream(idocstream & is) : is_(is) {}
+
+ /// Like std::istream::operator bool()
+ /// Do not convert is_ implicitly to bool, since that is forbidden in C++11.
+ explicit operator bool() const { return s_.empty() ? !is_.fail() : true; }
+
+ /// change the encoding of the input stream to \p e (iconv name)
+ void setEncoding(std::string const & e);
+
+ // add to the list of characters to read before actually reading
+ // the stream
+ void putback(char_type c);
+
+ // add to the list of characters to read before actually reading
+ // the stream
+ void putback(docstring s);
+
+ /// Like std::istream::get()
+ iparserdocstream & get(char_type &c);
+
+ /// Like std::istream::good()
+ bool good() const { return s_.empty() ? is_.good() : true; }
+
+ /// Like std::istream::peek()
+ int_type peek() const { return s_.empty() ? is_.peek() : s_[0]; }
+private:
+ ///
+ idocstream & is_;
+ /// characters to read before actually reading the stream
+ docstring s_;
+};
+
/*!
* Actual parser class
*/
class Parser {
-
+ /// noncopyable
+ Parser(Parser const & p);
+ Parser & operator=(Parser const & p);
public:
///
- Parser(std::istream & is);
+ Parser(idocstream & is, std::string const & fixedenc);
///
Parser(std::string const & s);
///
~Parser();
+ /** forget already parsed next tokens and put the
+ * corresponding characters into the input stream for
+ * re-reading. Useful when changing catcodes. */
+ void deparse();
+
+ /// change the encoding of the input stream according to \p encoding
+ /// (latex name) and package \p package
+ bool setEncoding(std::string const & encoding, int const & package);
+ /// change the encoding of the input stream to \p encoding (iconv name)
+ bool setEncoding(std::string const & encoding);
+ /// get the current iconv encoding of the input stream
+ std::string getEncoding() const { return encoding_iconv_; }
+
+ ///
+ CatCode catcode(char_type c) const;
+ ///
+ void setCatcode(char c, CatCode cat);
+ /// set parser to normal or verbatim mode
+ void setCatcodes(cat_type t);
+
///
int lineno() const { return lineno_; }
///
void putback();
+ /// store current position
+ void pushPosition();
+ /// restore previous position
+ void popPosition();
+ /// forget last saved position
+ void dropPosition();
/// dump contents to screen
void dump() const;
+ /// Does an optional argument follow after the current token?
+ bool hasOpt();
///
typedef std::pair<bool, std::string> Arg;
/*!
* Get an argument enclosed by \p left and \p right.
+ * If \p allow_escaping is true, a right delimiter escaped by a
+ * backslash does not count as delimiter, but is included in the
+ * argument.
* \returns wether an argument was found in \p Arg.first and the
* argument in \p Arg.second. \see getArg().
*/
- Arg getFullArg(char left, char right);
+ Arg getFullArg(char left, char right, bool allow_escaping = true);
/*!
* Get an argument enclosed by \p left and \p right.
+ * If \p allow_escaping is true, a right delimiter escaped by a
+ * backslash does not count as delimiter, but is included in the
+ * argument.
* \returns the argument (without \p left and \p right) or the empty
* string if the next non-space token is not \p left. Use
* getFullArg() if you need to know wether there was an empty
* argument or no argument at all.
*/
- std::string getArg(char left, char right);
+ std::string getArg(char left, char right, bool allow_escaping = true);
/*!
- * \returns getFullArg('[', ']') including the brackets or the
- * empty string if there is no such argument.
+ * Like getOpt(), but distinguishes between a missing argument ""
+ * and an empty argument "[]".
*/
- std::string getFullOpt();
+ std::string getFullOpt(bool keepws = false);
/*!
* \returns getArg('[', ']') including the brackets or the
* empty string if there is no such argument.
+ * No whitespace is eaten if \p keepws is true and no optional
+ * argument exists. This is important if an optional argument is
+ * parsed that would go after a command in ERT: In this case the
+ * whitespace is needed to separate the ERT from the subsequent
+ * word. Without it, the ERT and the next word would be concatenated
+ * during .tex export, thus creating an invalid command.
*/
- std::string getOpt();
+ std::string getOpt(bool keepws = false);
/*!
* \returns getFullArg('(', ')') including the parentheses or the
* empty string if there is no such argument.
/*!
* \returns the contents of the environment \p name.
* <tt>\begin{name}</tt> must be parsed already, <tt>\end{name}</tt>
- * is parsed but not returned.
+ * is parsed but not returned. This parses nested environments properly.
*/
- std::string const verbatimEnvironment(std::string const & name);
- /*!
- * Returns the character of the current token and increments
- * the token position.
+ std::string const ertEnvironment(std::string const & name);
+ /*
+ * The same as ertEnvironment(std::string const & name) but
+ * \begin and \end commands inside the name environment are not parsed.
+ * This function is designed to parse verbatim environments.
+ */
+ std::string const plainEnvironment(std::string const & name);
+ /*
+ * Basically the same as plainEnvironment(std::string const & name) but
+ * instead of \begin and \end commands the parsing is started/stopped
+ * at given characters.
+ * This function is designed to parse verbatim commands.
*/
- char getChar();
+ std::string const plainCommand(char left, char right, std::string const & name);
+ /*
+ * Basically the same as plainEnvironment() but the parsing is
+ * stopped at string \p end_string. Contrary to the other
+ * methods, this uses proper catcode setting. This function is
+ * designed to parse verbatim environments and command. The
+ * intention is to eventually replace all of its siblings. the
+ * member \p first of the result tells whether the arg was
+ * found and the member \p second is the value. If \p
+ * allow_linebreak is false, then the parsing is limited to one line
+ */
+ Arg verbatimStuff(std::string const & end_string,
+ bool allow_linebreak = true);
+ /*
+ * \returns the contents of the environment \p name.
+ * <tt>\begin{name}</tt> must be parsed already,
+ * <tt>\end{name}</tt> is parsed but not returned. The string
+ * is parsed with proper verbatim catcodes and one newline is
+ * removed from head and tail of the string if applicable.
+ */
+ std::string const verbatimEnvironment(std::string const & end_string);
///
- void error(std::string const & msg);
- /// Parses one token from \p is
- void tokenize_one();
- /// Parses \p is into tokens
- void tokenize();
+ std::string verbatim_item();
///
- void push_back(Token const & t);
+ std::string verbatimOption();
+ ///
+ void error(std::string const & msg);
/// The previous token.
- Token const & prev_token() const;
+ Token const prev_token() const;
/// The current token.
- Token const & curr_token() const;
- /// The next token.
- Token const & next_token() const;
+ Token const curr_token() const;
+ /// The next token. Caution: If this is called, an encoding change is
+ /// only possible again after get_token() has been called.
+ Token const next_token();
+ /// The next but one token. Caution: If this is called, an encoding
+ /// change is only possible again after get_token() has been called
+ /// twice.
+ Token const next_next_token();
/// Make the next token current and return that.
- Token const & get_token();
+ Token const get_token();
/// \return whether the current token starts a new paragraph
- bool isParagraph() const;
+ bool isParagraph();
/// skips spaces (and comments if \p skip_comments is true)
- void skip_spaces(bool skip_comments = false);
+ /// \return whether whitespace was skipped (not comments)
+ bool skip_spaces(bool skip_comments = false);
/// puts back spaces (and comments if \p skip_comments is true)
void unskip_spaces(bool skip_comments = false);
- ///
- void lex(std::string const & s);
- ///
- bool good() const;
- ///
- std::string verbatim_item();
- ///
- std::string verbatimOption();
+ /// Is any further input pending()? This is not like
+ /// std::istream::good(), which returns true if all available input
+ /// was read, and the next attempt to read would return EOF.
+ bool good();
/// resets the parser to initial state
void reset();
- ///
- void setCatCode(char c, CatCode cat);
- ///
- CatCode getCatCode(char c) const;
private:
+ /// Setup catcode table
+ void catInit();
+ /// Parses one token from \p is
+ void tokenize_one();
+ ///
+ void push_back(Token const & t);
///
int lineno_;
///
std::vector<Token> tokens_;
///
- unsigned pos_;
+ size_t pos_;
+ ///
+ std::vector<unsigned> positions_;
+ ///
+ idocstringstream * iss_;
+ ///
+ iparserdocstream is_;
+ /// iconv name of the current encoding
+ std::string encoding_iconv_;
+ ///
+ CatCode theCatcode_[256];
+ ///
+ cat_type theCatcodesType_;
///
- std::istringstream * iss_;
+ cat_type curr_cat_;
///
- std::istream & is_;
+ bool fixed_enc_;
};