csvlint 0.0.1

Sign up to get free protection for your applications and to get access to all the features.
Files changed (53) hide show
  1. checksums.yaml +7 -0
  2. data/.coveralls.yml +1 -0
  3. data/.gitignore +22 -0
  4. data/.travis.yml +10 -0
  5. data/Gemfile +7 -0
  6. data/LICENSE.md +22 -0
  7. data/README.md +214 -0
  8. data/Rakefile +17 -0
  9. data/bin/create_schema +32 -0
  10. data/bin/csvlint +52 -0
  11. data/csvlint.gemspec +39 -0
  12. data/features/check_format.feature +46 -0
  13. data/features/csv_options.feature +35 -0
  14. data/features/fixtures/cr-line-endings.csv +1 -0
  15. data/features/fixtures/crlf-line-endings.csv +3 -0
  16. data/features/fixtures/inconsistent-line-endings.csv +2 -0
  17. data/features/fixtures/invalid-byte-sequence.csv +24 -0
  18. data/features/fixtures/lf-line-endings.csv +3 -0
  19. data/features/fixtures/spreadsheet.xls +0 -0
  20. data/features/fixtures/title-row.csv +4 -0
  21. data/features/fixtures/valid.csv +3 -0
  22. data/features/fixtures/windows-line-endings.csv +2 -0
  23. data/features/information.feature +22 -0
  24. data/features/parse_csv.feature +90 -0
  25. data/features/schema_validation.feature +63 -0
  26. data/features/sources.feature +18 -0
  27. data/features/step_definitions/csv_options_steps.rb +19 -0
  28. data/features/step_definitions/information_steps.rb +13 -0
  29. data/features/step_definitions/parse_csv_steps.rb +30 -0
  30. data/features/step_definitions/schema_validation_steps.rb +7 -0
  31. data/features/step_definitions/sources_steps.rb +7 -0
  32. data/features/step_definitions/validation_errors_steps.rb +43 -0
  33. data/features/step_definitions/validation_info_steps.rb +18 -0
  34. data/features/step_definitions/validation_warnings_steps.rb +46 -0
  35. data/features/support/env.rb +30 -0
  36. data/features/support/webmock.rb +1 -0
  37. data/features/validation_errors.feature +151 -0
  38. data/features/validation_info.feature +24 -0
  39. data/features/validation_warnings.feature +74 -0
  40. data/lib/csvlint.rb +13 -0
  41. data/lib/csvlint/error_collector.rb +43 -0
  42. data/lib/csvlint/error_message.rb +15 -0
  43. data/lib/csvlint/field.rb +102 -0
  44. data/lib/csvlint/schema.rb +69 -0
  45. data/lib/csvlint/types.rb +113 -0
  46. data/lib/csvlint/validate.rb +253 -0
  47. data/lib/csvlint/version.rb +3 -0
  48. data/lib/csvlint/wrapped_io.rb +39 -0
  49. data/spec/field_spec.rb +247 -0
  50. data/spec/schema_spec.rb +149 -0
  51. data/spec/spec_helper.rb +20 -0
  52. data/spec/validator_spec.rb +279 -0
  53. metadata +367 -0
@@ -0,0 +1,69 @@
1
+ require "set"
2
+
3
+ module Csvlint
4
+
5
+ class Schema
6
+
7
+ include Csvlint::ErrorCollector
8
+
9
+ attr_reader :uri, :fields, :title, :description
10
+
11
+ def initialize(uri, fields=[], title=nil, description=nil)
12
+ @uri = uri
13
+ @fields = fields
14
+ @title = title
15
+ @description = description
16
+ reset
17
+ end
18
+
19
+ def validate_header(header)
20
+ reset
21
+ header.each_with_index do |name,i|
22
+ build_warnings(:header_name, :schema, nil, i+1, name) if fields[i].name != name
23
+ end
24
+ return valid?
25
+ end
26
+
27
+ def validate_row(values, row=nil)
28
+ reset
29
+ if values.length < fields.length
30
+ fields[values.size..-1].each_with_index do |field, i|
31
+ build_warnings(:missing_column, :schema, row, values.size+i+1)
32
+ end
33
+ end
34
+ if values.length > fields.length
35
+ values[fields.size..-1].each_with_index do |data_column, i|
36
+ build_warnings(:extra_column, :schema, row, fields.size+i+1)
37
+ end
38
+ end
39
+
40
+ fields.each_with_index do |field,i|
41
+ value = values[i] || ""
42
+ result = field.validate_column(value, row, i+1)
43
+ @errors += fields[i].errors
44
+ @warnings += fields[i].warnings
45
+ end
46
+
47
+ return valid?
48
+ end
49
+
50
+ def Schema.from_json_table(uri, json)
51
+ fields = []
52
+ json["fields"].each do |field_desc|
53
+ fields << Csvlint::Field.new( field_desc["name"] , field_desc["constraints"],
54
+ field_desc["title"], field_desc["description"] )
55
+ end if json["fields"]
56
+ return Schema.new( uri , fields, json["title"], json["description"] )
57
+ end
58
+
59
+ def Schema.load_from_json_table(uri)
60
+ begin
61
+ json = JSON.parse( open(uri).read )
62
+ return Schema.from_json_table(uri,json)
63
+ rescue
64
+ return nil
65
+ end
66
+ end
67
+
68
+ end
69
+ end
@@ -0,0 +1,113 @@
1
+ require 'set'
2
+ require 'date'
3
+ require 'active_support/core_ext/date/conversions'
4
+ require 'active_support/core_ext/time/conversions'
5
+
6
+ module Csvlint
7
+
8
+ module Types
9
+
10
+ SIMPLE_FORMATS = {
11
+ 'string' => lambda { |value, constraints| value },
12
+ 'numeric' => lambda do |value, constraints|
13
+ begin
14
+ Integer value
15
+ rescue ArgumentError
16
+ Float value
17
+ end
18
+ end,
19
+ 'uri' => lambda do |value, constraints|
20
+ u = URI.parse value
21
+ raise ArgumentError unless u.kind_of?(URI::HTTP) || u.kind_of?(URI::HTTPS)
22
+ u
23
+ end
24
+ }
25
+
26
+ def self.date_format(klass = DateTime, value, type)
27
+ date = klass.strptime(value, klass::DATE_FORMATS[type])
28
+ raise ArgumentError unless date.to_formatted_s(type) == value
29
+ end
30
+
31
+ def self.included(base)
32
+ Time::DATE_FORMATS[:iso8601] = "%Y-%m-%dT%H:%M:%SZ"
33
+ Time::DATE_FORMATS[:hms] = "%H:%M:%S"
34
+
35
+ Date::DATE_FORMATS.each do |type|
36
+ SIMPLE_FORMATS["date_#{type.first}"] = lambda do |value, constraints|
37
+ date_format(Date, value, type.first)
38
+ end
39
+ end
40
+
41
+ Time::DATE_FORMATS.each do |type|
42
+ SIMPLE_FORMATS["dateTime_#{type.first}"] = lambda do |value, constraints|
43
+ date_format(Time, value, type.first)
44
+ end
45
+ end
46
+ end
47
+
48
+ TYPE_VALIDATIONS = {
49
+ 'http://www.w3.org/2001/XMLSchema#string' => SIMPLE_FORMATS['string'],
50
+ 'http://www.w3.org/2001/XMLSchema#int' => lambda { |value, constraints| Integer value },
51
+ 'http://www.w3.org/2001/XMLSchema#integer' => lambda { |value, constraints| Integer value },
52
+ 'http://www.w3.org/2001/XMLSchema#float' => lambda { |value, constraints| Float value },
53
+ 'http://www.w3.org/2001/XMLSchema#double' => lambda { |value, constraints| Float value },
54
+ 'http://www.w3.org/2001/XMLSchema#anyURI' => SIMPLE_FORMATS['uri'],
55
+ 'http://www.w3.org/2001/XMLSchema#boolean' => lambda do |value, constraints|
56
+ return true if ['true', '1'].include? value
57
+ return false if ['false', '0'].include? value
58
+ raise ArgumentError
59
+ end,
60
+ 'http://www.w3.org/2001/XMLSchema#nonPositiveInteger' => lambda do |value, constraints|
61
+ i = Integer value
62
+ raise ArgumentError unless i <= 0
63
+ i
64
+ end,
65
+ 'http://www.w3.org/2001/XMLSchema#negativeInteger' => lambda do |value, constraints|
66
+ i = Integer value
67
+ raise ArgumentError unless i < 0
68
+ i
69
+ end,
70
+ 'http://www.w3.org/2001/XMLSchema#nonNegativeInteger' => lambda do |value, constraints|
71
+ i = Integer value
72
+ raise ArgumentError unless i >= 0
73
+ i
74
+ end,
75
+ 'http://www.w3.org/2001/XMLSchema#positiveInteger' => lambda do |value, constraints|
76
+ i = Integer value
77
+ raise ArgumentError unless i > 0
78
+ i
79
+ end,
80
+ 'http://www.w3.org/2001/XMLSchema#dateTime' => lambda do |value, constraints|
81
+ date_pattern = constraints["datePattern"] || "%Y-%m-%dT%H:%M:%SZ"
82
+ d = DateTime.strptime(value, date_pattern)
83
+ raise ArgumentError unless d.strftime(date_pattern) == value
84
+ d
85
+ end,
86
+ 'http://www.w3.org/2001/XMLSchema#date' => lambda do |value, constraints|
87
+ date_pattern = constraints["datePattern"] || "%Y-%m-%d"
88
+ d = Date.strptime(value, date_pattern)
89
+ raise ArgumentError unless d.strftime(date_pattern) == value
90
+ d
91
+ end,
92
+ 'http://www.w3.org/2001/XMLSchema#time' => lambda do |value, constraints|
93
+ date_pattern = constraints["datePattern"] || "%H:%M:%S"
94
+ d = DateTime.strptime(value, date_pattern)
95
+ raise ArgumentError unless d.strftime(date_pattern) == value
96
+ d
97
+ end,
98
+ 'http://www.w3.org/2001/XMLSchema#gYear' => lambda do |value, constraints|
99
+ date_pattern = constraints["datePattern"] || "%Y"
100
+ d = Date.strptime(value, date_pattern)
101
+ raise ArgumentError unless d.strftime(date_pattern) == value
102
+ d
103
+ end,
104
+ 'http://www.w3.org/2001/XMLSchema#gYearMonth' => lambda do |value, constraints|
105
+ date_pattern = constraints["datePattern"] || "%Y-%m"
106
+ d = Date.strptime(value, date_pattern)
107
+ raise ArgumentError unless d.strftime(date_pattern) == value
108
+ d
109
+ end
110
+ }
111
+ end
112
+
113
+ end
@@ -0,0 +1,253 @@
1
+ require "open_uri_redirections"
2
+
3
+ module Csvlint
4
+
5
+ class Validator
6
+
7
+ include Csvlint::ErrorCollector
8
+ include Csvlint::Types
9
+
10
+ attr_reader :encoding, :content_type, :extension, :headers, :line_breaks, :dialect, :csv_header, :schema, :data
11
+
12
+ ERROR_MATCHERS = {
13
+ "Missing or stray quote" => :stray_quote,
14
+ "Illegal quoting" => :whitespace,
15
+ "Unclosed quoted field" => :unclosed_quote,
16
+ }
17
+
18
+ def initialize(source, dialect = nil, schema = nil)
19
+ @source = source
20
+ @formats = []
21
+ @schema = schema
22
+
23
+ @assumed_header = true
24
+ @supplied_dialect = dialect != nil
25
+
26
+ if dialect
27
+ @assumed_header = false
28
+ end
29
+
30
+ @dialect = dialect_defaults = {
31
+ "header" => true,
32
+ "delimiter" => ",",
33
+ "skipInitialSpace" => true,
34
+ "lineTerminator" => :auto,
35
+ "quoteChar" => '"'
36
+ }.merge(dialect || {})
37
+
38
+ @csv_header = @dialect["header"]
39
+
40
+ @csv_options = dialect_to_csv_options(@dialect)
41
+ @extension = parse_extension(source)
42
+ reset
43
+ validate
44
+ end
45
+
46
+ def validate
47
+ single_col = false
48
+ io = nil
49
+ begin
50
+ io = @source.respond_to?(:gets) ? @source : open(@source, :allow_redirections=>:all)
51
+ validate_metadata(io)
52
+ parse_csv(io)
53
+ unless @col_counts.inject(:+).nil?
54
+ build_warnings(:title_row, :structure) if @col_counts.first < (@col_counts.inject(:+) / @col_counts.count)
55
+ end
56
+ build_warnings(:check_options, :structure) if @expected_columns == 1
57
+ check_consistency
58
+ rescue OpenURI::HTTPError, Errno::ENOENT
59
+ build_errors(:not_found)
60
+ ensure
61
+ io.close if io && io.respond_to?(:close)
62
+ end
63
+ end
64
+
65
+ def validate_metadata(io)
66
+ @encoding = io.charset rescue nil
67
+ @content_type = io.content_type rescue nil
68
+ @headers = io.meta rescue nil
69
+ if @headers
70
+ if @headers["content-type"] =~ /header=(present|absent)/
71
+ @csv_header = true if $1 == "present"
72
+ @csv_header = false if $1 == "absent"
73
+ @assumed_header = false
74
+ end
75
+ if @headers["content-type"] !~ /charset=/
76
+ build_warnings(:no_encoding, :context)
77
+ else
78
+ build_warnings(:encoding, :context) if @encoding != "utf-8"
79
+ end
80
+ build_warnings(:no_content_type, :context) if @content_type == nil
81
+ build_warnings(:excel, :context) if @content_type == nil && @extension =~ /.xls(x)?/
82
+ build_errors(:wrong_content_type, :context) unless (@content_type && @content_type =~ /text\/csv/)
83
+
84
+ if @assumed_header && !@supplied_dialect && (@content_type == nil || @headers["content-type"] !~ /header=(present|absent)/ )
85
+ build_errors(:undeclared_header, :structure)
86
+ end
87
+
88
+ end
89
+ build_info_messages(:assumed_header, :structure) if @assumed_header
90
+ end
91
+
92
+ def parse_csv(io)
93
+ @expected_columns = 0
94
+ current_line = 0
95
+ reported_invalid_encoding = false
96
+ @col_counts = []
97
+
98
+ @csv_options[:encoding] = @encoding
99
+
100
+ begin
101
+ wrapper = WrappedIO.new( io )
102
+ csv = CSV.new( wrapper, @csv_options )
103
+ @data = []
104
+ @line_breaks = csv.row_sep
105
+ if @line_breaks != "\r\n"
106
+ build_info_messages(:nonrfc_line_breaks, :structure)
107
+ end
108
+ row = nil
109
+ loop do
110
+ current_line = current_line + 1
111
+ begin
112
+ row = csv.shift
113
+ @data << row
114
+ wrapper.finished
115
+ if row
116
+ if header? && current_line == 1
117
+ row = row.reject {|r| r.blank? }
118
+ validate_header(row)
119
+ @col_counts << row.count
120
+ else
121
+ build_formats(row, current_line)
122
+ @col_counts << row.reject {|r| r.blank? }.count
123
+ @expected_columns = row.count unless @expected_columns != 0
124
+
125
+ build_errors(:blank_rows, :structure, current_line, nil, wrapper.line) if row.reject{ |c| c.nil? || c.empty? }.count == 0
126
+
127
+ if @schema
128
+ @schema.validate_row(row, current_line)
129
+ @errors += @schema.errors
130
+ @warnings += @schema.warnings
131
+ else
132
+ build_errors(:ragged_rows, :structure, current_line, nil, wrapper.line) if !row.empty? && row.count != @expected_columns
133
+ end
134
+
135
+ end
136
+ else
137
+ break
138
+ end
139
+ rescue CSV::MalformedCSVError => e
140
+ wrapper.finished
141
+ type = fetch_error(e)
142
+ if type == :stray_quote && !wrapper.line.match(csv.row_sep)
143
+ build_errors(:line_breaks, :structure)
144
+ else
145
+ build_errors(type, :structure, current_line, nil, wrapper.line)
146
+ end
147
+ end
148
+ end
149
+ rescue ArgumentError => ae
150
+ wrapper.finished
151
+ build_errors(:invalid_encoding, :structure, current_line, wrapper.line) unless reported_invalid_encoding
152
+ reported_invalid_encoding = true
153
+ end
154
+ end
155
+
156
+ def validate_header(header)
157
+ names = Set.new
158
+ header.each_with_index do |name,i|
159
+ build_warnings(:empty_column_name, :schema, nil, i+1) if name == ""
160
+ if names.include?(name)
161
+ build_warnings(:duplicate_column_name, :schema, nil, i+1)
162
+ else
163
+ names << name
164
+ end
165
+ end
166
+ if @schema
167
+ @schema.validate_header(header)
168
+ @errors += @schema.errors
169
+ @warnings += @schema.warnings
170
+ end
171
+ return valid?
172
+ end
173
+
174
+ def header?
175
+ return @csv_header
176
+ end
177
+
178
+ def fetch_error(error)
179
+ e = error.message.match(/^([a-z ]+) (i|o)n line ([0-9]+)\.?$/i)
180
+ message = e[1] rescue nil
181
+ ERROR_MATCHERS.fetch(message, :unknown_error)
182
+ end
183
+
184
+ def dialect_to_csv_options(dialect)
185
+ skipinitialspace = dialect["skipInitialSpace"] || true
186
+ delimiter = dialect["delimiter"]
187
+ delimiter = delimiter + " " if !skipinitialspace
188
+ return {
189
+ :col_sep => delimiter,
190
+ :row_sep => dialect["lineTerminator"],
191
+ :quote_char => dialect["quoteChar"],
192
+ :skip_blanks => false
193
+ }
194
+ end
195
+
196
+ def build_formats(row, line)
197
+ row.each_with_index do |col, i|
198
+ next if col.blank?
199
+ @formats[i] ||= []
200
+
201
+ SIMPLE_FORMATS.each do |type, lambda|
202
+ begin
203
+ lambda.call(col, {})
204
+ @format = type
205
+ rescue => e
206
+ nil
207
+ end
208
+ end
209
+
210
+ @formats[i] << @format
211
+ end
212
+ end
213
+
214
+ def check_consistency
215
+ percentages = []
216
+
217
+ formats = SIMPLE_FORMATS.map {|type, lambda| type }
218
+
219
+ formats.each do |type, regex|
220
+ @formats.count.times do |i|
221
+ percentages[i] ||= {}
222
+ unless @formats[i].nil?
223
+ percentages[i][type] = @formats[i].grep(/^#{type}$/).count.to_f / @formats[i].count.to_f
224
+ end
225
+ end
226
+ end
227
+
228
+ percentages.each_with_index do |col, i|
229
+ next if col.values.blank?
230
+ build_warnings(:inconsistent_values, :schema, nil, i+1) if col.values.max < 0.9
231
+ end
232
+ end
233
+
234
+ private
235
+
236
+ def parse_extension(source)
237
+ case source
238
+ when File
239
+ return File.extname( source.path )
240
+ when IO
241
+ return ""
242
+ when StringIO
243
+ return ""
244
+ when Tempfile
245
+ return ""
246
+ else
247
+ parsed = URI.parse(source)
248
+ File.extname(parsed.path)
249
+ end
250
+ end
251
+
252
+ end
253
+ end
@@ -0,0 +1,3 @@
1
+ module Csvlint
2
+ VERSION = "0.0.1"
3
+ end
@@ -0,0 +1,39 @@
1
+ module Csvlint
2
+ class WrappedIO
3
+ def initialize(io)
4
+ @io = io
5
+ @line = ""
6
+ end
7
+
8
+ def gets(*args)
9
+ if args.count == 1 && args[0].is_a?(String)
10
+ delim = args[0]
11
+ @line = "" if @new_line
12
+ s = @io.gets(delim)
13
+ if s != nil
14
+ @line << s
15
+ end
16
+ return s
17
+ else
18
+ @io.gets(*args)
19
+ end
20
+ end
21
+
22
+ def eof?
23
+ @io.eof?
24
+ end
25
+
26
+ def finished
27
+ @new_line = true
28
+ end
29
+
30
+ def line
31
+ @line
32
+ end
33
+
34
+ def method_missing(method, *args)
35
+ @io.send(method, *args)
36
+ end
37
+
38
+ end
39
+ end