Sha256: d7e5a710c94e9157ccaec28a35d7f73283f6d5802d0cde3d5e889a6a09b0884b
Contents?: true
Size: 1.94 KB
Versions: 25
Compression:
Stored size: 1.94 KB
Contents
class Licensee
class LevenshteinMatcher < Matcher
# Return the first potential license that is more similar than the confidence threshold
def match
@match ||= potential_licenses.find do |license|
similarity(license) >= Licensee::CONFIDENCE_THRESHOLD
end
end
# Sort all licenses, in decending order, by difference in length to the file
# Difference in lengths cannot exceed the file's length * the confidence threshold / 100
def potential_licenses
@potential_licenses ||= begin
Licensee.licenses.select { |license| length_delta(license) <= max_delta }.sort_by { |l| length_delta(l) }
end
end
# Calculate the difference between the file length and a given license's length
def length_delta(license)
(file_length - license.body_normalized.length).abs
end
# Maximum possible difference between file length and license length
# for a license to be a potential license to be matched
def max_delta
@max_delta ||= (file_length * (Licensee::CONFIDENCE_THRESHOLD.to_f / 100.to_f ))
end
# Confidence that the matched license is a match
def confidence
@confidence ||= match ? similarity(match) : 0
end
private
# Length of the file, normalized to strip whitespace
def file_length
@file_length ||= file.content_normalized.length.to_f
end
# Calculate percent changed between file and potential license
def similarity(license)
100 * (file_length - distance(license)) / file_length
end
# Calculate the levenshtein distance between file and license
# Note: We used content/body normalized because white space and capitalization
# isn't legally significant in this context. Fewer characters lets levenshtein
# work faster. As long as they both undergo the same transformation, should match.
def distance(license)
Levenshtein.distance(license.body_normalized, file.content_normalized).to_f
end
end
end
Version data entries
25 entries across 25 versions & 1 rubygems