proper timezone awareness and plugin parser hooks

Parser uses method compilation technique namely to a method not a proc
This commit is contained in:
Adam Meehan
2010-09-16 22:35:38 +10:00
parent 9ddd150b2f
commit 5d495505d9
7 changed files with 813 additions and 60 deletions

View File

@@ -4,13 +4,10 @@ require 'active_support/core_ext/hash/except'
require 'active_support/core_ext/string/conversions'
require 'active_support/core_ext/date/acts_like'
require 'active_support/core_ext/date/conversions'
require 'active_support/core_ext/date/zones'
require 'active_support/core_ext/time/acts_like'
require 'active_support/core_ext/time/conversions'
require 'active_support/core_ext/time/zones'
require 'active_support/core_ext/date_time/acts_like'
require 'active_support/core_ext/date_time/conversions'
require 'active_support/core_ext/date_time/zones'
module ValidatesTimeliness
autoload :VERSION, 'validates_timeliness/version'
@@ -19,6 +16,14 @@ module ValidatesTimeliness
mattr_accessor :extend_orms
@@extend_orms = [ defined?(ActiveRecord) && :active_record ].compact
# User the plugin date/time parser which is stricter and extendable
mattr_accessor :use_plugin_parser
@@use_plugin_parser = false
# Default timezone
mattr_accessor :default_timezone
@@default_timezone = defined?(ActiveRecord) ? ActiveRecord::Base.default_timezone : :utc
# Set the dummy date part for a time type values.
mattr_accessor :dummy_date_for_time_type
@@dummy_date_for_time_type = [ 2000, 1, 1 ]
@@ -41,6 +46,7 @@ module ValidatesTimeliness
end
end
require 'validates_timeliness/parser'
require 'validates_timeliness/conversion'
require 'validates_timeliness/validator'
require 'validates_timeliness/helper_methods'

View File

@@ -8,7 +8,7 @@ module ValidatesTimeliness
when :date
value.to_date
when :datetime
value.to_time.in_time_zone
value.to_time
end
rescue
nil
@@ -16,29 +16,34 @@ module ValidatesTimeliness
def dummy_time(value)
time = if value.acts_like?(:time)
value = value.in_time_zone
[value.hour, value.min, value.sec]
else
[0,0,0]
end
Time.local(*(ValidatesTimeliness.dummy_date_for_time_type + time))
Time.send(ValidatesTimeliness.default_timezone, *(ValidatesTimeliness.dummy_date_for_time_type + time))
end
def evaluate_option_value(value, record)
def evaluate_option_value(value, record, timezone_aware=false)
case value
when Time, Date
when Time
timezone_aware ? value.in_time_zone : value
when Date
value
when String
value.to_time(:local)
if ValidatesTimeliness.use_plugin_parser
ValidatesTimeliness::Parser.parse(value, :datetime, :timezone_aware => timezone_aware, :strict => false)
else
timezone_aware ? Time.zone.parse(value) : value.to_time(ValidatesTimeliness.default_timezone)
end
when Symbol
if !record.respond_to?(value) && restriction_shorthand?(value)
ValidatesTimeliness.restriction_shorthand_symbols[value].call
else
evaluate_option_value(record.send(value), record)
evaluate_option_value(record.send(value), record, timezone_aware)
end
when Proc
result = value.arity > 0 ? value.call(record) : value.call
evaluate_option_value(result, record)
evaluate_option_value(result, record, timezone_aware)
else
value
end

View File

@@ -0,0 +1,407 @@
require 'date'
module ValidatesTimeliness
# A date and time parsing library which allows you to add custom formats using
# simple predefined tokens. This makes it much easier to catalogue and customize
# the formats rather than dealing directly with regular expressions.
#
# Formats can be added or removed to customize the set of valid date or time
# string values.
#
class Parser
cattr_accessor :time_formats,
:date_formats,
:datetime_formats,
:time_expressions,
:date_expressions,
:datetime_expressions,
:format_tokens,
:format_proc_args
# Set the threshold value for a two digit year to be considered last century
#
# Default: 30
#
# Example:
# year = '29' is considered 2029
# year = '30' is considered 1930
#
cattr_accessor :ambiguous_year_threshold
self.ambiguous_year_threshold = 30
# Set the dummy date part for a time type value. Should be an array of 3 values
# being year, month and day in that order.
#
# Default: [ 2000, 1, 1 ] same as ActiveRecord
#
cattr_accessor :dummy_date_for_time_type
self.dummy_date_for_time_type = [ 2000, 1, 1 ]
# Format tokens:
# y = year
# m = month
# d = day
# h = hour
# n = minute
# s = second
# u = micro-seconds
# ampm = meridian (am or pm) with or without dots (e.g. am, a.m, or a.m.)
# _ = optional space
# tz = Timezone abbreviation (e.g. UTC, GMT, PST, EST)
# zo = Timezone offset (e.g. +10:00, -08:00, +1000)
#
# All other characters are considered literal. You can embed regexp in the
# format but no gurantees that it will remain intact. If you avoid the use
# of any token characters and regexp dots or backslashes as special characters
# in the regexp, it may well work as expected. For special characters use
# POSIX character clsses for safety.
#
# Repeating tokens:
# x = 1 or 2 digits for unit (e.g. 'h' means an hour can be '9' or '09')
# xx = 2 digits exactly for unit (e.g. 'hh' means an hour can only be '09')
#
# Special Cases:
# yy = 2 or 4 digit year
# yyyy = exactly 4 digit year
# mmm = month long name (e.g. 'Jul' or 'July')
# ddd = Day name of 3 to 9 letters (e.g. Wed or Wednesday)
# u = microseconds matches 1 to 6 digits
#
# Any other invalid combination of repeating tokens will be swallowed up
# by the next lowest length valid repeating token (e.g. yyy will be
# replaced with yy)
@@time_formats = [
'hh:nn:ss',
'hh-nn-ss',
'h:nn',
'h.nn',
'h nn',
'h-nn',
'h:nn_ampm',
'h.nn_ampm',
'h nn_ampm',
'h-nn_ampm',
'h_ampm'
]
@@date_formats = [
'yyyy-mm-dd',
'yyyy/mm/dd',
'yyyy.mm.dd',
'm/d/yy',
'd/m/yy',
'm\d\yy',
'd\m\yy',
'd-m-yy',
'd.m.yy',
'd mmm yy'
]
@@datetime_formats = [
'yyyy-mm-dd hh:nn:ss',
'yyyy-mm-dd h:nn',
'yyyy-mm-dd h:nn_ampm',
'yyyy-mm-dd hh:nn:ss.u',
'm/d/yy h:nn:ss',
'm/d/yy h:nn_ampm',
'm/d/yy h:nn',
'd/m/yy hh:nn:ss',
'd/m/yy h:nn_ampm',
'd/m/yy h:nn',
'ddd, dd mmm yyyy hh:nn:ss (zo|tz)', # RFC 822
'ddd mmm d hh:nn:ss zo yyyy', # Ruby time string
'yyyy-mm-ddThh:nn:ssZ', # iso 8601 without zone offset
'yyyy-mm-ddThh:nn:sszo' # iso 8601 with zone offset
]
# All tokens available for format construction. The token array is made of
# token regexp, validation regexp and key for format proc mapping if any.
# If the token needs no format proc arg then the validation regexp should
# not have a capturing group, as all captured groups are passed to the
# format proc.
#
# The token regexp should only use a capture group if 'look-behind' anchor
# is required. The first capture group will be considered a literal and put
# into the validation regexp string as-is. This is a hack.
@@format_tokens = [
{ 'ddd' => [ /d{3}/, '\w{3,9}' ] },
{ 'dd' => [ /d{2}/, '\d{2}', :day ] },
{ 'd' => [ /d/, '\d{1,2}', :day ] },
{ 'ampm' => [ /ampm/, '[aApP]\.?[mM]\.?', :meridian ] },
{ 'mmm' => [ /m{3}/, '\w{3,9}', :month ] },
{ 'mm' => [ /m{2}/, '\d{2}', :month ] },
{ 'm' => [ /m{1}/, '\d{1,2}', :month ] },
{ 'yyyy' => [ /y{4}/, '\d{4}', :year ] },
{ 'yy' => [ /y{2}/, '\d{4}|\d{2}', :year ] },
{ 'hh' => [ /h{2}/, '\d{2}', :hour ] },
{ 'h' => [ /h{1}/, '\d{1,2}', :hour ] },
{ 'nn' => [ /n{2}/, '\d{2}', :min ] },
{ 'n' => [ /n{1}/, '\d{1,2}', :min ] },
{ 'ss' => [ /s{2}/, '\d{2}', :sec ] },
{ 's' => [ /s{1}/, '\d{1,2}', :sec ] },
{ 'u' => [ /u{1}/, '\d{1,6}', :usec ] },
{ 'zo' => [ /zo/, '[+-]\d{2}:?\d{2}', :offset ] },
{ 'tz' => [ /tz/, '[A-Z]{1,4}' ] },
{ '_' => [ /_/, '\s?' ] }
]
# Arguments which will be passed to the format proc if matched in the
# time string. The key must be the key from the format tokens. The array
# consists of the arry position of the arg, the arg name, and the code to
# place in the time array slot. The position can be nil which means the arg
# won't be placed in the array.
#
# The code can be used to manipulate the arg value if required, otherwise
# should just be the arg name.
#
@@format_proc_args = {
:year => [0, 'y', 'unambiguous_year(y)'],
:month => [1, 'm', 'month_index(m)'],
:day => [2, 'd', 'd'],
:hour => [3, 'h', 'full_hour(h, md ||= nil)'],
:min => [4, 'n', 'n'],
:sec => [5, 's', 's'],
:usec => [6, 'u', 'microseconds(u)'],
:offset => [7, 'z', 'offset_in_seconds(z)'],
:meridian => [nil, 'md', nil]
}
@@type_wrapper = {
:date => [/\A/, nil],
:time => [nil , /\Z/],
:datetime => [/\A/, /\Z/]
}
class << self
def compile_format_expressions
@@time_expressions = compile_formats(@@time_formats)
@@date_expressions = compile_formats(@@date_formats)
@@datetime_expressions = compile_formats(@@datetime_formats)
end
def parse(raw_value, type, options={})
return nil if raw_value.blank?
return raw_value if raw_value.acts_like?(:time) || raw_value.acts_like?(:date)
time_array = _parse(raw_value, type, options.reverse_merge(:strict => true))
return nil if time_array.nil?
if type == :date
Date.new(*time_array[0..2]) rescue nil
else
make_time(time_array[0..7], options[:timezone_aware])
end
end
def make_time(time_array, timezone_aware=false)
# Enforce strict date part validity which Time class does not
return nil unless Date.valid_civil?(*time_array[0..2])
if timezone_aware
Time.zone.local(*time_array)
else
Time.time_with_datetime_fallback(ValidatesTimeliness.default_timezone, *time_array)
end
rescue ArgumentError, TypeError
nil
end
# Loop through format expressions for type and call the format method on a match.
# Allow pre or post match strings to exist if strict is false. Otherwise wrap
# regexp in start and end anchors.
#
# Returns time array if matches a format, nil otherwise.
#
def _parse(string, type, options={})
options.reverse_merge!(:strict => true)
sets = if options[:format]
options[:strict] = true
[ send("#{type}_expressions").assoc(options[:format]) ]
else
expression_set(type, string)
end
set = sets.find do |format, regexp|
string =~ wrap_regexp(regexp, type, options[:strict])
end
if set
last = options[:include_offset] ? 8 : 7
values = send(:"format_#{set[0]}", *$~[1..last])
values[0..2] = ValidatesTimeliness.dummy_date_for_time_type if type == :time
return values
end
end
# Delete formats of specified type. Error raised if format not found.
def remove_formats(type, *remove_formats)
remove_formats.each do |format|
unless self.send("#{type}_formats").delete(format)
raise "Format #{format} not found in #{type} formats"
end
end
compile_format_expressions
end
# Adds new formats. Must specify format type and can specify a :before
# option to nominate which format the new formats should be inserted in
# front on to take higher precedence.
# Error is raised if format already exists or if :before format is not found.
def add_formats(type, *add_formats)
formats = self.send("#{type}_formats")
options = {}
options = add_formats.pop if add_formats.last.is_a?(Hash)
before = options[:before]
raise "Format for :before option #{format} was not found." if before && !formats.include?(before)
add_formats.each do |format|
raise "Format #{format} is already included in #{type} formats" if formats.include?(format)
index = before ? formats.index(before) : -1
formats.insert(index, format)
end
compile_format_expressions
end
# Removes formats where the 1 or 2 digit month comes first, to eliminate
# formats which are ambiguous with the European style of day then month.
# The mmm token is ignored as its not ambigous.
def remove_us_formats
us_format_regexp = /\Am{1,2}[^m]/
date_formats.reject! { |format| us_format_regexp =~ format }
datetime_formats.reject! { |format| us_format_regexp =~ format }
compile_format_expressions
end
def full_hour(hour, meridian)
hour = hour.to_i
return hour if meridian.nil?
if meridian.delete('.').downcase == 'am'
hour == 12 ? 0 : hour
else
hour == 12 ? hour : hour + 12
end
end
def unambiguous_year(year)
if year.length <= 2
century = Time.now.year.to_s[0..1].to_i
century -= 1 if year.to_i >= ambiguous_year_threshold
year = "#{century}#{year.rjust(2,'0')}"
end
year.to_i
end
def month_index(month)
return month.to_i if month.to_i.nonzero?
abbr_month_names.index(month.capitalize) || month_names.index(month.capitalize)
end
def month_names
I18n.t('date.month_names')
end
def abbr_month_names
I18n.t('date.abbr_month_names')
end
def microseconds(usec)
(".#{usec}".to_f * 1_000_000).to_i
end
def offset_in_seconds(offset)
sign = offset =~ /^-/ ? -1 : 1
parts = offset.scan(/\d\d/).map {|p| p.to_f }
parts[1] = parts[1].to_f / 60
(parts[0] + parts[1]) * sign * 3600
end
private
# Generate regular expression from format string
def generate_format_expression(string_format)
format = string_format.dup
format.gsub!(/([\.\\])/, '\\\\\1') # escapes dots and backslashes
found_tokens, token_order = [], []
format_tokens.each do |token|
token_regexp, regexp_str, arg_key = *token.values.first
if format.gsub!(token_regexp, "%<#{found_tokens.size}>")
regexp_str = "(#{regexp_str})" if arg_key
found_tokens << [regexp_str, arg_key]
end
end
format.scan(/%<(\d)>/).each {|token_index|
token_index = token_index.first
token = found_tokens[token_index.to_i]
format.gsub!("%<#{token_index}>", token[0])
token_order << token[1]
}
compile_format_method(token_order.compact, string_format)
Regexp.new(format)
rescue
raise "The following format regular expression failed to compile: #{format}\n from format #{string_format}."
end
# Compiles a format method which maps the regexp capture groups to method
# arguments based on order captured. A time array is built using the values
# in the position indicated by the first element of the proc arg array.
#
def compile_format_method(order, name)
values = [nil] * 7
args = []
order.each do |part|
proc_arg = format_proc_args[part]
args << proc_arg[1]
values[proc_arg[0]] = proc_arg[2] if proc_arg[0]
end
class_eval <<-DEF
class << self
define_method(:"format_#{name}") do |#{args.join(',')}|
[#{values.map {|i| i || 'nil' }.join(',')}].map {|i| i.is_a?(Float) ? i : i.to_i }
end
end
DEF
end
def compile_formats(formats)
formats.map { |format| [ format, generate_format_expression(format) ] }
end
# Pick expression set and combine date and datetimes for
# datetime attributes to allow date string as datetime
def expression_set(type, string)
case type
when :date
date_expressions
when :time
time_expressions
when :datetime
# gives a speed-up for date string as datetime attributes
if string.length < 11
date_expressions + datetime_expressions
else
datetime_expressions + date_expressions
end
end
end
def wrap_regexp(regexp, type, strict=false)
type = strict ? :datetime : type
/#{@@type_wrapper[type][0]}#{regexp}#{@@type_wrapper[type][1]}/
end
end
end
end
ValidatesTimeliness::Parser.compile_format_expressions

View File

@@ -41,13 +41,14 @@ module ValidatesTimeliness
raw_value = attribute_raw_value(record, attr_name) || value
return if (@allow_nil && raw_value.nil?) || (@allow_blank && raw_value.blank?)
timezone_aware = record.class.timeliness_attribute_timezone_aware?(attr_name)
value = type_cast(value)
return record.errors.add(attr_name, :"invalid_#{@type}") if value.blank?
@restrictions_to_check.each do |restriction|
begin
restriction_value = type_cast(evaluate_option_value(options[restriction], record))
restriction_value = type_cast(evaluate_option_value(options[restriction], record, timezone_aware))
unless value.send(RESTRICTIONS[restriction], restriction_value)
return record.errors.add(attr_name, restriction, :message => options[:"#{restriction}_message"], :restriction => format_error_value(restriction_value))
end