require 'http_client_patch/include_client'
require 'httpclient'
require 'nokogiri'
require 'time'
require 'uri'

require 'summon'
require 'summon/transport/headers'

# Search engine for Serial Solutions Summon
#
# Docs:
# http://api.summon.serialssolutions.com/help/api/search
# http://api.summon.serialssolutions.com/help/api/search/fields
#
# Also Summon user guide at:
# http://support.serialssolutions.com/app/answers/detail/a_id/3906
#
# Can try out searches and see XML results with an auth login in their
# interface: http://api.summon.serialssolutions.com/help/api/search/example?s.q=query
#
# An example user-facing Summon UI, useful for figuring out available
# facets and facet values, or trying out searches:
# http://ncsu.summon.serialssolutions.com/

#
# == Functionality notes
#
# * for pagination, underlying summon API only supports 'page', not 'start'
#   style, if you pass in 'start' style it will be 'rounded' to containing 'page'.
#
# == Required config params
# [access_id]   supplied by SerSol for your account
# [secret_key]  supplied by SerSol for your account
#
# == Optional custom config params
#
# [fixed_params]
#     Fixed SerSol query param literals to send with every search.
#     Value is a HASH, of keys and either single values or arrays
#     of values. For instance, to exclude certain content types
#     from all search results, in config:
#         :fixed_params =>
#           {"s.fvf" => ["ContentType,Web Resource,true","ContentType,Reference,true","ContentType,eBook,true"] }
#     Note that values are NOT URI escaped in config, code will take care
#     of that for you. You could also fix "s.role" to 'authenticated' using
#     this mechanism, if you restrict all access to your app to authenticated
#     affiliated users.
# [highlighting]
#     Default true, ask SerSol for query-in-context highlighting in
#     title and snippets field. If true you WILL get HTML with <b> tags
#     in your titles -- and snippets available in ResultItems. They
#     will be used by stnadard display logic as summary unless you set
#     for_display.prefer_abstract_as_summary = true
# [use_summon_openurl] default false. If true, will use OpenURL kev context
#     object passed back by summon to generate openurls, instead of creating
#     one ourself from individual data elements. summon openurl is decent,
#     but currently includes highlighting tags in title elements. Also note
#     it includes DC-type openurls, which we don't currently generate ourselves.
# [lang]  Sent to summon as "s.l" param, see http://api.summon.serialssolutions.com/help/api/search/parameters/language
#         default nil. You may want to set to "en".
#
#
# == Custom search params
#
# [:auth] Pass in `:auth => true` (or "true") to send headers to summon
# indicating an authorized user, for full search results.
# [:summon_params]  Hash of key/value pairs to pass directly to summon. Just like
#                  fixed_params in configuration, but per-search. Can be
#                  used to directly trigger functionality not covered by
#                  the bento_search adapter. Values can be arrays where summon
#                  keys are repeatable.
# [:peer_reviewed_only]   Set to boolean true or string 'true', to restrict
#                         results to peer-reviewed only (as identified by Summon)
# [:online_only]      Limit to only items marked 'with fulltext online' by Summon.
#                     Just a convenience shortcut for setting s.fvf including IsFullText,true
#
# [:pubyear_start]
# [:pubyear_end]          Date range limiting, pass in custom search args,
#                         one or both of pubyear_start and pubyear_end
#                         #to_i will be called on it, so can be string.
#                         .search(:query => "foo", :pubyear_start => 2000)
#
# == Custom response data
#
# * result_item.custom_data["summon.original_data"] has a Hash of the decoded JSON
#   for the item, directly as returned by Summon API.
#
# == Tech notes
# We did not choose to use the summon ruby gem in general, we wanted more control
# than it offered (ability to use HTTPClient persistent connections, MultiJson
# for json parsing, etc).
#
# However, we DO use that gem specifically for constructing authentication
# headers how summon wants it, see class at
# https://github.com/summon/summon.rb/blob/master/lib/summon/transport/headers.rb
#
# Language provided only in language_str not language_code, all that API gives
# us. We could try to reverse lookup from ISO code labels later if we want.
class BentoSearch::SummonEngine
  include BentoSearch::SearchEngine


  # Can't change http timeout in config, because we keep an http
  # client at class-wide level, and config is not class-wide.
  # Change this 'constant' if you want to change it, I guess.
  #
  # Summon is pretty fast, we think a 4.5 second timeout should be
  # be plenty. May be adjusted from experience.
  HttpTimeout = 4.5
  extend HTTPClientPatch::IncludeClient
  include_http_client do |client|
    client.connect_timeout = client.send_timeout = client.receive_timeout = HttpTimeout
  end

  include ActionView::Helpers::OutputSafetyHelper # for safe_join

  # Originally we used $$BENTO_HL_START$$ etc with dollar
  # signs, but the dollar signs trigger a weird bug in summon
  # where end tokens are missing from output.
  @@hl_start_token = "__BENTO_HL_START__"
  @@hl_end_token = "__BENTO_HL_END__"

  def search_implementation(args)
    uri, headers = construct_request(args)

    Rails.logger.debug("SummonEngine request URL: #{uri}")

    results = BentoSearch::Results.new

    hash, response, exception = nil
    begin
      response = http_client.get(uri, nil, headers)
      hash = MultiJson.load( response.body )
    rescue TimeoutError, HTTPClient::ConfigurationError, HTTPClient::BadResponseError, MultiJson::DecodeError, Nokogiri::SyntaxError => e
      exception = e
    end
    # handle some errors
    if (response.nil? || hash.nil? || exception ||
      (! HTTP::Status.successful? response.status))
      results.error ||= {}
      results.error[:exception] = e
      results.error[:status] = response.status if response

      return results
    end

    results.total_items = hash["recordCount"]

    hash["documents"].each do |doc_hash|
      item = BentoSearch::ResultItem.new

      item.custom_data["summon.original_data"] = doc_hash

      item.unique_id      = first_if_present doc_hash["ID"]

      item.title          = format_title(doc_hash)

      item.link           = doc_hash["link"]
      # Don't understand difference between hasFullText and
      # isFullTextHit. ??. We'll use hasFullText for now, that's
      # the documented one.
      item.link_is_fulltext = doc_hash["hasFullText"]

      if configuration.use_summon_openurl
        item.openurl_kev_co = doc_hash["openUrl"] # Summon conveniently gives us pre-made OpenURL
      end

      item.journal_title  = first_if_present doc_hash["PublicationTitle"]
      item.issn           = first_if_present doc_hash["ISSN"]
      item.isbn           = first_if_present doc_hash["ISBN"]
      item.doi            = first_if_present doc_hash["DOI"]

      item.start_page     = first_if_present doc_hash["StartPage"]
      item.end_page       = first_if_present doc_hash["EndPage"]

      if (pubdate = first_if_present doc_hash["PublicationDate_xml"])
        item.year         = pubdate["year"]
      end
      item.volume         = first_if_present doc_hash["Volume"]
      item.issue          = first_if_present doc_hash["Issue"]

      if (pub = first_if_present doc_hash["Publisher_xml"])
        item.publisher    = pub["name"]
      end

      # if it's a dissertation, put the school in the 'publisher' field.
      # if we don't have one otherwise.
      if (! item.publisher) && (school = first_if_present doc_hash["DissertationSchool_xml"])
        item.publisher    = school["name"]
      end

      (doc_hash["Author_xml"] || []).each do |auth_hash|
        a = BentoSearch::Author.new

        a.first           = name_normalize auth_hash["givenname"]
        a.last            = name_normalize auth_hash["surname"]
        a.middle          = name_normalize auth_hash["middlename"]

        a.display         = name_normalize auth_hash["fullname"]

        item.authors << a unless a.empty?
      end

      item.format         = normalize_content_type( first_if_present doc_hash["ContentType"] )
      if doc_hash["ContentType"]
        item.format_str     = doc_hash["ContentType"].join(", ")
      end

      item.language_str   = first_if_present doc_hash["Language"]

      item.abstract       = first_if_present doc_hash["Abstract"]

      # Just straight snippets
      if doc_hash["Snippet"]
        item.snippets = doc_hash["Snippet"].collect {|s| handle_highlighting(s)}
      end

      results << item
    end


    return results
  end

  # Looks up a record by SerSol Id, engine.get( item.unique_id ) should return
  # item.
  #
  # Returns a single BentoSearch::ResultItem, or raises BentoSearch::NotFound,
  # BentoSearch::TooManyFound, or other unspecified exception.
  def get(id)
    # "ID" is an internal search field for Summon, not listed in our
    # own search_field_definitions, but it works.
    results = search(id, :search_field => "ID")

    raise BentoSearch::NotFound.new("ID: #{id}") if results.length == 0
    raise BentoSearch::TooManyFound.new("ID: #{id}") if results.length == 0
    raise (results.error[:exception] || Exception.new(error.inspect)) if results.failed?

    return results.first
  end

  def first_if_present(array)
    array ? array.first : nil
  end


  # Normalize Summon Content-Type to our standardized
  # list.
  #
  # This ends up losing useful distinctions Summon makes, however.
  def normalize_content_type(summon_type)
    case summon_type
    when "Journal Article", "Book Review", "Trade Publication Article", "Newspaper Article" then "Article"
    when "Audio Recording", "Music Recording", "Spoken Word Recording" then "AudioObject"
    when "Book", "eBook", "Book / eBook" then "Book"
    when "Book Chapter" then :book_item
    when "Conference Proceeding" then :conference_paper
    when "Dissertation" then :dissertation
    when "Journal", "Newsletter", "Newspaper" then :serial
    when "Photograph" then "Photograph"
    when "Report", "Technical Report"   then "Report"
    when "Video Recording", "Film" then "VideoObject"
    when "Web Resource" then "WebPage"
    when "Computer File", "Data Set" then "SoftwareApplication"
    else nil
    end
  end

  def name_normalize(str)

    return nil if str.blank?

    str = str.strip

    return nil if str.blank? || str =~ /^[,:.]*$/

    return str
  end


  # returns two element array: [uri, headers]
  #
  # uri, headers = construct_request(args)
  def construct_request(args)
    # Query params in a hash with array values, becuase easiest
    # to generate auth headers that way. Value is array of values that
    # are NOT URI-encoded yet.
    query_params = Hash.new {|h, k| h[k] = [] }

    # Add in fixed_params from config, and summon_params from search, if any.

    direct_params = (configuration.fixed_params || {}).merge( args[:summon_params] || {} )

    direct_params.each_pair do |key, value|
      [value].flatten.each do |v|
        query_params[key] << v
      end
    end

    if args[:per_page]
      query_params["s.ps"] = args[:per_page]
    end
    if args[:page]
      query_params["s.pn"] = args[:page]
    end

    if args[:search_field]
      query_params['s.q'] = "#{args[:search_field]}:(#{summon_escape(args[:query])})"
    else
      query_params['s.q'] = summon_escape( args[:query] )
    end

    if (args[:sort] &&
        (defn = self.sort_definitions[args[:sort]]) &&
        (literal = defn[:implementation]))
      query_params['s.sort'] =  literal
    end

    if args[:auth] == true
      query_params['s.role'] = "authenticated"
    end

    if [true, "true"].include? args[:peer_reviewed_only]
      query_params['s.fvf'] ||= []
      query_params['s.fvf'] << "IsPeerReviewed,true"
    end

    if [true, "true"].include? args[:online_only]
      query_params["s.fvf"] ||= []
      query_params['s.fvf'] << "IsFullText,true"
    end

    # Summon uses "*" for open ended range endpoint
    if args[:pubyear_start] || args[:pubyear_end]
      from = args[:pubyear_start].to_i
      from = "*" if from == 0

      to = args[:pubyear_end].to_i
      to = "*" if to == 0

      query_params["s.rf"] ||= []
      query_params["s.rf"] << "PublicationDate,#{from}:#{to}"
    end

    if configuration.highlighting
      query_params['s.hs'] = @@hl_start_token
      query_params['s.he'] = @@hl_end_token
    else
      query_params['s.hl'] = "false"
    end

    if configuration.lang
      query_params["s.l"] = configuration.lang
    end


    headers = Summon::Transport::Headers.new(
      :access_id => configuration.access_id,
      :secret_key => configuration.secret_key,
      :accept => "json",
      :params => query_params,
      :url => configuration.base_url
      )


    query_string = query_params.keys.collect do |key|
      [query_params[key]].flatten.collect do |value|
        "#{CGI.escape(key.to_s)}=#{CGI.escape(value.to_s)}"
      end
    end.flatten.join("&")

    uri = "#{configuration.base_url}?#{query_string}"

    return [uri, headers]
  end


  # Escapes special chars for Summon. Not entirely clear what
  # we have to escape where (or double escape sometimes?), but
  # we're just going to do a straight backslash escape of special
  # chars.
  #
  # Does NOT do URI-escaping, that's a different step.
  def summon_escape(string)
    # replace with backslash followed by original matched thing,
    # need to double backslash for ruby string literal makes
    # this ridiculously confusing, sorry. Block form of gsub
    # is the only thing that keeps it from being impossible.
    #
    # Do NOT escape double quotes, let people use them for
    # phrases!
    #
    # While docs suggest you have to double-slash escape hyphens,
    # in fact doing so ruins ID: search, so we don't.
    string.gsub(/([+&|!\(\){}\[\]^~*?\\:])/) do |match|
      "\\#{$1}"
    end
  end

  # If summon has put snippet highlighting tokens
  # in a field, we need to HTML escape the literal values,
  # while still using the highlighting tokens to put
  # HTML tags around highlighted terms.
  def handle_highlighting( str, options = {} )
    BentoSearch::Util.handle_highlight_tags(
      str,
      :start_tag => @@hl_start_token,
      :end_tag => @@hl_end_token,
      :enabled => configuration.highlighting,
      :strip => options[:strip]
      )
  end

  # combine title and subtitle into one title,
  def format_title(doc_hash)
    title          = first_if_present  doc_hash["Title"]
    subtitle       = first_if_present doc_hash["Subtitle"]

    if subtitle.present?
      title = "#{title}: #{subtitle}"
    end

    return handle_highlighting( title )
  end

  def self.required_configuration
    [:access_id, :secret_key]
  end

  def self.default_configuration
    {
      :base_url => "http://api.summon.serialssolutions.com/2.0.0/search",
      :highlighting => true,
      :use_summon_openurl => false
    }
  end

  def max_per_page
    200
  end

  # Summon actually only supports relevancy sort, and pub year asc or desc.
  # we just expose relevance and pub year desc here.
  def sort_definitions
    # implementation includes literal sersol value, but not yet
    # uri escaped, that'll happen at a later code point.
    {
      "relevance" => {:implementation => nil}, # default
      "date_desc" => {:implementation => "PublicationDate:desc"}

    }
  end

  # Summon offers many more search fields than this. This is a subset
  # listed here. See http://api.summon.serialssolutions.com/help/api/search/fields
  # although those docs may not be up to date.
  #
  # The AuthorCombined, TitleCombined, and SubjectCombined indexes
  # aren't even listed in the docs, but they are real. I think.
  def search_field_definitions
      {
        nil                   => {:semantic => :general},
        "AuthorCombined"      => {:semantic => :author},
        "TitleCombined"       => {:semantic => :title},
        # SubjectTerms does not include TemporalSubjectTerms
        # or Keywords, sorry.
        "SubjectTerms"        => {:semantic => :subject},
        # ISBN and ISSN do not include seperate EISSN and EISBN
        # fields, sorry.
        "ISBN"                => {:semantic => :isbn},
        "ISSN"                => {:semantic => :issn},
        "OCLC"                => {:semantic => :oclcnum},
        "PublicationSeriesTitle" => {:semantic => :publication_title }
      }
  end



end