# text.rb : A simple backend to deal with basic text files.
# Copyright (C) 2006 Vincent Fourmond

# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# (at your option) any later version.

# This program 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 General Public License for more details.

# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301 USA



require 'Dobjects/Dvector'
require 'Dobjects/Function'

# For separated sets
require 'stringio'

module CTioga2

  module Data

    # A module for easy use of NaN in operations
    module NaN
      NaN = 0.0/0.0
      def nan
        return NaN
      end
    end

    module Backends

      class TextBackend < Backend

        # A constant holding a relation extension -> command to
        # decompress (to be fed to sprintf with the filename as argument)
        UNCOMPRESSORS = {
          ".gz" => "gunzip -c %s",
          ".bz2" => "bunzip2 -c %s",
          ".lzma" => "unlzma -c %s",
          ".lz" => "unlzma -c %s",
          ".xz" => "unxz -c %s",
        }

        include Dobjects

        describe 'text', 'Text format', <<EOD
This backend reads text files in a format close to the one understood
by gnuplot and the like. More specifically, it reads text files organized
in columns.

The column separator is specified by the @separator@ option or using
the {command: text-separator} command; it is a {type: regexp}

By default, the {command: text} backend only loads the column 1 as X
and 2 as Y, but that can be changed either by specifiying a default
column spec using either the @default_column_spec@ option or by adding
a specification directly after the file name.

EOD

        # Inherit the baseline handling, can be useful !
        # inherit_parameters :base_line
        
        param_accessor :skip, 'skip', "Skip lines", 'integer', 
        "Number of lines to be skipped at the beginning of the file"

        param_accessor :default_column_spec, 'col', 
        "Default column specification", 'text', 
        "Which columns to use when the @1:2 syntax is not used"

        param_accessor :split, 'split', "Split into subsets", 'boolean', 
        "If true, splits files into subsets on blank/non number lines"


        param_accessor :separator, 'separator', "Data columns separator", 
        'regexp', 
        "The columns separator. Defaults to /\s+/"

        param_accessor :param_regex, 'parameters', "Parameters parsing", 
        'regexp', 
        "Regular expression for extracting parameters from a file. Defaults to nil (ie nothing)"

        param_accessor :header_line_regex, 'header-line', 
        'Header line regular expression', 
        'regexp', 
        "Regular expression indicating the header line (containing column names) (default /^##/"
        
        def initialize
          @dummy = nil
          @current = nil   
          # Current is the name of the last file used. Necessary for '' specs.
          @current_data = nil       # The data of the last file used.
          @skip = 0
          @included_modules = [NaN]    # to make sure we give them to
          # Dvector.compute_formula
          @default_column_spec = "1:2"

          @separator = /\s+/

          # We don't split data by default.
          @split = false

          @param_regex = nil

          @header_line_regex = /^\#\#\s*/

          super()

          # Override Backend's cache - for now.
          @cache = {}               # A cache file_name -> data

          @param_cache = {}     # Same thing as cache, but for parameters

          @headers_cache = {}   # Same thing as cache, but for header
                                # lines.

        end

        def extend(mod)
          super
          @included_modules << mod
        end

        # Expands specifications into few sets. This function will separate the
        # set into a file spec and a col spec. Within the col spec, the 2##6
        # keyword is used to expand to 2,3,4,5,6. 2## followed by a non-digit
        # expands to 2,...,last column in the file. For now, the expansions
        # stops on the first occurence found, and the second form doesn't
        # work yet. But soon...
        def expand_sets(spec)
          if m = /(\d+)##(\D|$)/.match(spec)
            a = m[1].to_i 
            trail = m[2]
            b = read_file(spec)
            b = (b.length - 1) 
            ret = []
            a.upto(b) do |i|
              ret << m.pre_match + i.to_s + trail + m.post_match
            end
            return ret
          else
            m = Dir::glob(spec)
            if m.size > 0
              return m
            else
              return super
            end
          end
        end


        protected


        # Returns a IO object suitable to acquire data from it for
        # the given _file_, which can be one of the following:
        # * a real file name
        # * a compressed file name
        # * a pipe command.
        def get_io_object(file)
          if file == "-"
            return $stdin
          elsif file =~ /(.*?)\|\s*$/ # A pipe
            return IO.popen($1)
          elsif not File.readable?(file)
            # Try to find a compressed version
            for ext,method in UNCOMPRESSORS
              if File.readable? "#{file}#{ext}"
                info { "Using compressed file #{file}#{ext} in stead of #{file}" }
                return IO.popen(method % "#{file}#{ext}")
              end
            end
          else 
            for ext, method in UNCOMPRESSORS
              if file =~ /#{ext}$/ 
                info { "Taking file #{file} as a compressed file" }
                return IO.popen(method % file)
              end
            end
          end
          return File::open(file)
        end

        # A line is invalid if it is blank or starts
        # neither with a digit nor +, - or .
        #
        # Maybe to be improved later.
        InvalidLineRE = /^\s*$|^\s*[^\d+.\s-]+/

        # Returns a string corresponding to the given _set_ of the
        # given _io_ object.
        #
        # Sets are 1-based.
        def get_set_string(io, set)
          cur_set = 1
          last_line_is_invalid = true
          str = ""
          line_number = 0
          while line = io.gets
            line_number += 1
            if line =~ InvalidLineRE
              debug { "Found invalid line at #{line_number}" }
              if ! last_line_is_invalid
                # We begin a new set.
                cur_set += 1
                debug { "Found set #{cur_set} at line #{line_number}" }
                if(cur_set > set)
                  return str
                end
              end
              last_line_is_invalid = true
            else
              last_line_is_invalid = false
              if cur_set == set
                str += line
              end
            end
          end
          return str
        end

        # Returns an IO object corresponding to the given file.
        def get_io_set(file)
          if not @split
            return get_io_object(file)
          else
            file =~ /(.*?)(?:#(\d+))?$/; # ; to make ruby-mode indent correctly.
            filename = $1
            if $2
              set = $2.to_i
            else
              set = 1
            end
            debug { "Trying to get set #{set} from file '#{filename}'" }
            str = get_set_string(get_io_object(filename), set)
            return StringIO.new(str)
          end
        end

        undef :param_regex=
        # A proper writer for @param_regex
        def param_regex=(val)
          if val.is_a? Regexp
            @param_regex = val
          elsif val =~ /([^\\]|^)\(/     # Has capturing groups
            @param_regex = /#{val}/
          else                  # Treat as separator
            @param_regex = /(\S+)\s*#{val}\s*(\S+)/
          end
        end

        # Turns an array of comments into a hash[param] -> value
        def parse_parameters(comments)
          ret = {}
          for line in comments
            if line =~ @param_regex
              ret[$1] = $2.to_f
            end
          end
          return ret
        end

        # Turns an array of comments into a hash column name -> column
        # number (1-based)
        def parse_header_line(comments)
          for line in comments
            if line =~ @header_line_regex
              colnames = line.gsub(@header_line_regex,'').split(@separator)
              i = 1
              ret = {}
              for n in colnames
                ret[n] = i
                i += 1
              end
              return ret
            end
          end
          return {}
        end

        # Reads data from a file. If needed, extract the file from the
        # columns specification.
        #
        # \todo the cache really should include things such as time of
        # last modification and various parameters that influence the
        # reading of the file, and the parameters read from the file
        # using #parse_parameters
        #
        # \todo There should be a real global handling of meta-data
        # extracted from files, so that they could be included for
        # instance in the automatic labels ? (and we could have fun
        # improving this one ?)
        #
        # @todo There should be a way to read pure text columns and
        # use them somehow, to annotate the output ? This should be
        # implemented at the Tioga level, though (both for reading, in
        # fancy_read, and for using hover stuff)
        #
        # \warning This needs Tioga r561
        def read_file(file)
          if file =~ /(.*)@.*/
            file = $1
          end
          name = file               # As file will be modified.
          if ! @cache.key?(file)    # Read the file if it is not cached.
            comments = []
            fancy_read_options = {'index_col' => true,
              'skip_first' => @skip,
              'sep' => @separator,
              'comment_out' => comments
            }
            io_set = get_io_set(file)
            debug { "Fancy read '#{file}', options #{fancy_read_options.inspect}" }
            @cache[name] = Dvector.fancy_read(io_set, nil, fancy_read_options)
            if @param_regex
              # Now parsing params
              @param_cache[name] = parse_parameters(comments)
              info { "Read #{@param_cache[name].size} parameters from #{name}" }
              debug { "Parameters read: #{@param_cache[name].inspect}" }
            end
            if @header_line_regex
              @headers_cache[name] = parse_header_line(comments)
              info { "Read #{@headers_cache[name].size} column names from #{name}" }
              debug { "Got: #{@headers_cache[name].inspect}" }
            end
          end
          ## @todo These are not very satisfying; ideally, the data
          ## information should be embedded into @cache[name] rather
          ## than as external variables. Well...
          @current_parameters = @param_cache[name]
          @current_header = @headers_cache[name]
          return @cache[name]
        end


        # This is called by the architecture to get the data. It
        # splits the set name into filename@cols, reads the file if
        # necessary and calls get_data
        def query_dataset(set)
          if set =~ /(.*)@(.*)/
            col_spec = $2
            file = $1
          else
            col_spec = @default_column_spec
            file = set
          end
          if file.length > 0
            @current_data = read_file(file)
            @current = file
          end

          # Wether we need or not to compute formulas:
          if col_spec =~ /\$/
            compute_formulas = true
          else
            compute_formulas = false
          end
          
          return Dataset.dataset_from_spec(set, col_spec) do |col|
            get_data_column(col, compute_formulas, 
                            @current_parameters, @current_header)
          end
        end

        # Gets the data corresponding to the given column. If
        # _compute_formulas_ is true, the column specification is
        # taken to be a formula (in the spirit of gnuplot's)
        def get_data_column(column, compute_formulas = false, 
                            parameters = nil, header = nil)
          if compute_formulas
            formula = Utils::parse_formula(column, parameters, header)
            debug { "Using formula #{formula} for column spec: #{column}" }
            return Ruby.compute_formula(formula, 
                                        @current_data,
                                        @included_modules)
          else
            if @current_data[column.to_i]
              return @current_data[column.to_i].dup
            else
              raise "Cannot find column number #{column.to_i} -- maybe you got the column separator wrong ?"
            end
          end
        end

#         # Turns a target => values specification into something usable as
#         # error bars, that is :xmin, :xmax and the like hashes. The rules
#         # are the following:
#         # * ?min/?max are passed on directly;
#         # * ?e(abs) are transformed into ?min = ? - ?eabs, ?max = ? + ?eabs
#         # * ?eu(p/?ed(own) are transformed respectively into ? +/- ?...
#         # * ?er(el) become ?min = ?*(1 - ?erel, ?max = ?(1 + ?erel)
#         # * ?erup/?erdown follow the same pattern...
#         def compute_error_bars(values)
#           target = {}
#           for key in values.keys
#             case key.to_s
#             when /^[xy](min|max)?$/
#               target[key] = values[key].dup # Just to make sure.
#             when /^(.)e(a(bs?)?)?$/
#               target["#{$1}min".to_sym] = values[$1.to_sym] - values[key]
#               target["#{$1}max".to_sym] = values[$1.to_sym] + values[key]
#             when /^(.)eu(p)?$/
#               target["#{$1}max".to_sym] = values[$1.to_sym] + values[key]
#             when /^(.)ed(o(wn?)?)?$/
#               target["#{$1}min".to_sym] = values[$1.to_sym] - values[key]
#             when /^(.)er(el?)?$/
#               target["#{$1}min".to_sym] = values[$1.to_sym] * 
#                 (values[key].neg + 1)
#               target["#{$1}max".to_sym] = values[$1.to_sym] * 
#                 (values[key] + 1)
#             when /^(.)erd(o(wn?)?)?$/
#               target["#{$1}min".to_sym] = values[$1.to_sym] * 
#                 (values[key].neg + 1)
#             when /^(.)erup?$/
#               target["#{$1}max".to_sym] = values[$1.to_sym] * 
#                 (values[key] + 1)
#             else
#               warn "Somehow, the target specification #{key} " +
#                 "didn't make it through"
#             end
#           end
#           return target
#         end

      end

      
    end

  end
end