// file      : libbutl/manifest-parser.mxx -*- C++ -*-
// copyright : Copyright (c) 2014-2019 Code Synthesis Ltd
// license   : MIT; see accompanying LICENSE file

#ifndef __cpp_modules
#pragma once
#endif

// C includes.

#ifndef __cpp_lib_modules
#include <string>
#include <iosfwd>
#include <cstdint>    // uint64_t
#include <utility>    // pair, move()
#include <stdexcept>  // runtime_error
#include <functional>
#endif

// Other includes.

#ifdef __cpp_modules
export module butl.manifest_parser;
#ifdef __cpp_lib_modules
import std.core;
import std.io;
#endif
import butl.char_scanner;
import butl.manifest_types;
#else
#include <libbutl/char-scanner.mxx>
#include <libbutl/manifest-types.mxx>
#endif

#include <libbutl/export.hxx>

LIBBUTL_MODEXPORT namespace butl
{
  class LIBBUTL_SYMEXPORT manifest_parsing: public std::runtime_error
  {
  public:
    manifest_parsing (const std::string& name,
                      std::uint64_t line,
                      std::uint64_t column,
                      const std::string& description);

    manifest_parsing (const std::string& description);

    std::string name;
    std::uint64_t line;
    std::uint64_t column;
    std::string description;
  };

  class LIBBUTL_SYMEXPORT manifest_parser: protected butl::char_scanner
  {
  public:
    // The filter, if specified, is called by next() prior to returning the
    // pair to the caller. If the filter returns false, then the pair is
    // discarded.
    //
    // Note that the filter should handle the end-of-manifest pairs (see
    // below) carefully, so next() doesn't end up with an infinite cycle.
    //
    using filter_function = bool (manifest_name_value&);

    manifest_parser (std::istream& is,
                     const std::string& name,
                     std::function<filter_function> filter = {})
      : char_scanner (is), name_ (name), filter_ (std::move (filter)) {}

    const std::string&
    name () const {return name_;}

    // The first returned pair is special "start-of-manifest" with
    // empty name and value being the format version: {"", "<ver>"}.
    // After that we have a sequence of ordinary pairs which are
    // the manifest. At the end of the manifest we have the special
    // "end-of-manifest" pair with empty name and value: {"", ""}.
    // After that we can either get another start-of-manifest pair
    // (in which case the whole sequence repeats from the beginning)
    // or we get another end-of-manifest pair which signals the end
    // of stream (aka EOF). To put it another way, the parse sequence
    // always has the following form:
    //
    // ({"", "<ver>"} {"<name>", "<value>"}* {"", ""})* {"", ""}
    //
    manifest_name_value
    next ();

    // Split the manifest value, optionally followed by ';' character and a
    // comment into the value/comment pair. Note that ';' characters in the
    // value must be escaped by the backslash.
    //
    static std::pair<std::string, std::string>
    split_comment (const std::string&);

  private:
    void
    parse_next (manifest_name_value&);

    void
    parse_name (manifest_name_value&);

    void
    parse_value (manifest_name_value&);

    // Skip spaces and return the first peeked non-space character and the
    // starting position of the line it belongs to. If the later is not
    // available (skipped spaces are all in the middle of a line, we are at
    // eos, etc.), then fallback to the first peeked character position.
    //
    std::pair<xchar, std::uint64_t>
    skip_spaces ();

  private:
    const std::string name_;
    const std::function<filter_function> filter_;

    enum {start, body, end} s_ = start;
    std::string version_; // Current format version.
  };
}

#include <libbutl/manifest-parser.ixx>