Merge pull request JuliaLang#12799 from omus/dates_extensibility

Made DateFormat Extensible

base/dates/adjusters.jl

@@ -39,6 +39,7 @@ function lastdayofyear(dt::Date)
  y,m,d = yearmonthday(dt)
  return Date(UTD(value(dt)+daysinyear(y)-dayofyear(y,m,d)))
 end
+lastdayofyear(dt::DateTime) = DateTime(lastdayofyear(Date(dt)))
 
 @vectorize_1arg TimeType firstdayofyear
 @vectorize_1arg TimeType lastdayofyear

base/dates/io.jl

@@ -34,123 +34,161 @@ const MONTHTOVALUE = Dict{UTF8String,Dict{UTF8String,Int}}("english"=>english)
 const MONTHTOVALUEABBR = Dict{UTF8String,Dict{UTF8String,Int}}("english"=>abbrenglish)
 
 # Date/DateTime Parsing
-abstract Slot{P<:AbstractTime}
+abstract Slot{T<:Any}
 
-immutable DelimitedSlot{P<:AbstractTime} <: Slot{P}
- i::Int
- period::Type{P}
+immutable DelimitedSlot{T<:Any} <: Slot{T}
+ parser::Type{T}
+ letter::Char
  width::Int
- option::Int
- locale::AbstractString
+ transition::Union{Regex,AbstractString}
 end
 
-immutable FixedWidthSlot{P<:AbstractTime} <: Slot{P}
- i::Int
- period::Type{P}
+immutable FixedWidthSlot{T<:Any} <: Slot{T}
+ parser::Type{T}
+ letter::Char
  width::Int
- option::Int
- locale::AbstractString
 end
 
 immutable DateFormat
  slots::Array{Slot,1}
- begtran # optional transition from the start of a string to the 1st slot
- trans #trans[i] == how to transition FROM slots[i] TO slots[i+1]
+ prefix::AbstractString # optional transition from the start of a string to the 1st slot
+ locale::AbstractString
+end
+
+abstract DayOfWeekSlot
+
+# Slot rules translate letters into types. Note that
+# list of rules can be extended.
+immutable SlotRule
+ rules::Array{Type}
 end
+const SLOT_RULE = SlotRule(Array{Type}(256))
 
-immutable DayOfWeekSlot <: AbstractTime end
+getindex(collection::SlotRule, key::Char) = collection.rules[Int(key)]
+setindex!(collection::SlotRule, value::Type, key::Char) = collection.rules[Int(key)] = value
+keys(c::SlotRule) = map(Char, filter(el -> isdefined(c.rules, el), eachindex(c.rules)))
 
-duplicates(slots) = any(map(x->count(y->x.period==y.period,slots),slots) .> 1)
+SLOT_RULE['y'] = Year
+SLOT_RULE['m'] = Month
+SLOT_RULE['u'] = Month
+SLOT_RULE['U'] = Month
+SLOT_RULE['e'] = DayOfWeekSlot
+SLOT_RULE['E'] = DayOfWeekSlot
+SLOT_RULE['d'] = Day
+SLOT_RULE['H'] = Hour
+SLOT_RULE['M'] = Minute
+SLOT_RULE['S'] = Second
+SLOT_RULE['s'] = Millisecond
+
+duplicates(slots) = any(map(x->count(y->x.parser==y.parser,slots),slots) .> 1)
 
 function DateFormat(f::AbstractString,locale::AbstractString="english")
  slots = Slot[]
- trans = []
- begtran = match(r"^.*?(?=[ymuUdHMSsEe])",f).match
- ss = split(f,r"^.*?(?=[ymuUdHMSsEe])")
- s = split(begtran == "" ? ss[1] : ss[2],r"[^ymuUdHMSsEe]+|(?<=([ymuUdHMSsEe])(?!\1))")
- for (i,k) in enumerate(s)
- k == "" && break
- tran = i >= endof(s) ? r"$" : match(Regex("(?<=$(s[i])).*(?=$(s[i+1]))"),f).match
- slot = tran == "" ? FixedWidthSlot : DelimitedSlot
- width = length(k)
- typ = 'E' in k ? DayOfWeekSlot : 'e' in k ? DayOfWeekSlot :
- 'y' in k ? Year : 'm' in k ? Month :
- 'u' in k ? Month : 'U' in k ? Month :
- 'd' in k ? Day : 'H' in k ? Hour :
- 'M' in k ? Minute : 'S' in k ? Second : Millisecond
- option = 'E' in k ? 2 : 'e' in k ? 1 :
- 'U' in k ? 2 : 'u' in k ? 1 : 0
- push!(slots,slot(i,typ,width,option,locale))
- push!(trans,tran)
+ prefix = ""
+ params = ()
+ last_offset = 1
+
+ letters = join(keys(SLOT_RULE), "")
+ for m in eachmatch(Regex("([\\Q$letters\\E])\\1*"), f)
+ letter = f[m.offset]
+ typ = SLOT_RULE[letter]
+
+ width = length(m.match)
+ tran = f[last_offset:m.offset-1]
+
+ if isempty(params)
+ prefix = tran
+ else
+ slot = tran == "" ? FixedWidthSlot(params...) : DelimitedSlot(params..., tran)
+ push!(slots,slot)
+ end
+
+ params = (typ,letter,width)
+ last_offset = m.offset + width
+ end
+
+ tran = last_offset > endof(f) ? r"(?=\s|$)" : f[last_offset:end]
+ if !isempty(params)
+ slot = tran == "" ? FixedWidthSlot(params...) : DelimitedSlot(params..., tran)
+ push!(slots,slot)
  end
+
  duplicates(slots) && throw(ArgumentError("Two separate periods of the same type detected"))
- return DateFormat(slots,begtran,trans)
+ return DateFormat(slots,prefix,locale)
 end
 
 const SLOTERROR = ArgumentError("Non-digit character encountered")
-slotparse(slot,x) = !ismatch(r"[^0-9\s]",x) ? slot.period(x) : throw(SLOTERROR)
-function slotparse(slot::Slot{Month},x)
- if slot.option == 0
+slotparse(slot,x,locale) = !ismatch(r"[^0-9\s]",x) ? slot.parser(x) : throw(SLOTERROR)
+function slotparse(slot::Slot{Month},x,locale)
+ if slot.letter == 'm'
  ismatch(r"[^0-9\s]",x) ? throw(SLOTERROR) : return Month(x)
- elseif slot.option == 1
- return Month(MONTHTOVALUEABBR[slot.locale][lowercase(x)])
+ elseif slot.letter == 'u'
+ return Month(MONTHTOVALUEABBR[locale][lowercase(x)])
  else
- return Month(MONTHTOVALUE[slot.locale][lowercase(x)])
+ return Month(MONTHTOVALUE[locale][lowercase(x)])
  end
 end
-slotparse(slot::Slot{Millisecond},x) = !ismatch(r"[^0-9\s]",x) ? slot.period(Base.parse(Float64,"."*x)*1000.0) : throw(SLOTERROR)
-slotparse(slot::Slot{DayOfWeekSlot},x) = nothing
+slotparse(slot::Slot{Millisecond},x,locale) = !ismatch(r"[^0-9\s]",x) ? slot.parser(Base.parse(Float64,"."*x)*1000.0) : throw(SLOTERROR)
+slotparse(slot::Slot{DayOfWeekSlot},x,locale) = nothing
 
-function getslot(x,slot::DelimitedSlot,df,cursor)
- endind = first(search(x,df.trans[slot.i],nextind(x,cursor)))
+function getslot(x,slot::DelimitedSlot,locale,cursor)
+ endind = first(search(x,slot.transition,nextind(x,cursor)))
  if endind == 0 # we didn't find the next delimiter
  s = x[cursor:end]
- return (endof(x)+1, isdigit(s) ? slotparse(slot,s) : default(slot.period))
+ index = endof(x)+1
+ else
+ s = x[cursor:(endind-1)]
+ index = nextind(x,endind)
  end
- return nextind(x,endind), slotparse(slot,x[cursor:(endind-1)])
+ return index, slotparse(slot,s,locale)
 end
-getslot(x,slot,df,cursor) = (cursor+slot.width, slotparse(slot,x[cursor:(cursor+slot.width-1)]))
+getslot(x,slot,locale,cursor) = (cursor+slot.width, slotparse(slot,x[cursor:(cursor+slot.width-1)], locale))
 
 function parse(x::AbstractString,df::DateFormat)
- x = strip(replace(x, r"#.*$", ""))
- x = replace(x,df.begtran,"")
+ x = strip(x)
+ startswith(x, df.prefix) && (x = replace(x, df.prefix, "", 1))
  isempty(x) && throw(ArgumentError("Cannot parse empty format string"))
- (typeof(df.slots[1]) <: DelimitedSlot && first(search(x,df.trans[1])) == 0) && throw(ArgumentError("Delimiter mismatch. Couldn't find first delimiter, \"$(df.trans[1])\", in date string"))
+ if isa(df.slots[1], DelimitedSlot) && first(search(x,df.slots[1].transition)) == 0
+ throw(ArgumentError("Delimiter mismatch. Couldn't find first delimiter, \"$(df.slots[1].transition)\", in date string"))
+ end
  periods = Period[]
+ extra = []
  cursor = 1
  for slot in df.slots
- cursor, pe = getslot(x,slot,df,cursor)
- pe !== nothing && push!(periods,pe)
+ cursor, pe = getslot(x,slot,df.locale,cursor)
+ pe != nothing && (isa(pe,Period) ? push!(periods,pe) : push!(extra,pe))
  cursor > endof(x) && break
  end
- return sort!(periods,rev=true,lt=periodisless)
+ return vcat(sort!(periods,rev=true,lt=periodisless), extra)
 end
 
-slotformat(slot::Slot{Year},dt) = lpad(string(value(slot.period(dt))),slot.width,"0")[(end-slot.width+1):end]
-slotformat(slot,dt) = lpad(string(value(slot.period(dt))),slot.width,"0")
-function slotformat(slot::Slot{Month},dt)
- if slot.option == 0
+slotformat(slot::Slot{Year},dt,locale) = lpad(string(value(slot.parser(dt))),slot.width,"0")[(end-slot.width+1):end]
+slotformat(slot,dt,locale) = lpad(string(value(slot.parser(dt))),slot.width,"0")
+function slotformat(slot::Slot{Month},dt,locale)
+ if slot.letter == 'm'
  return lpad(month(dt),slot.width,"0")
- elseif slot.option == 1
- return VALUETOMONTHABBR[slot.locale][month(dt)]
+ elseif slot.letter == 'u'
+ return VALUETOMONTHABBR[locale][month(dt)]
  else
- return VALUETOMONTH[slot.locale][month(dt)]
+ return VALUETOMONTH[locale][month(dt)]
  end
 end
-function slotformat(slot::Slot{DayOfWeekSlot},dt)
- if slot.option == 1
- return VALUETODAYOFWEEKABBR[slot.locale][dayofweek(dt)]
- else # == 2
- return VALUETODAYOFWEEK[slot.locale][dayofweek(dt)]
+function slotformat(slot::Slot{DayOfWeekSlot},dt,locale)
+ if slot.letter == 'e'
+ return VALUETODAYOFWEEKABBR[locale][dayofweek(dt)]
+ else # == 'E'
+ return VALUETODAYOFWEEK[locale][dayofweek(dt)]
  end
 end
-slotformat(slot::Slot{Millisecond},dt) = rpad(string(millisecond(dt)/1000.0)[3:end], slot.width, "0")
+slotformat(slot::Slot{Millisecond},dt,locale) = rpad(string(millisecond(dt)/1000.0)[3:end], slot.width, "0")
 
 function format(dt::TimeType,df::DateFormat)
- f = ""
+ f = df.prefix
  for slot in df.slots
- f *= slotformat(slot,dt)
- f *= typeof(df.trans[slot.i]) <: Regex ? "" : df.trans[slot.i]
+ f *= slotformat(slot,dt,df.locale)
+ if isa(slot, DelimitedSlot)
+ f *= isa(slot.transition, AbstractString) ? slot.transition : ""
+ end
  end
  return f
 end

test/dates/adjusters.jl

@@ -183,6 +183,17 @@ end
 @test Dates.lastdayofquarter(Dates.DateTime(2014,8,2)) == Dates.DateTime(2014,9,30)
 @test Dates.lastdayofquarter(Dates.DateTime(2014,12,2)) == Dates.DateTime(2014,12,31)
 
+first = Dates.Date(2014,1,1)
+last = Dates.Date(2014,12,31)
+for i = 0:364
+ dt = first + Dates.Day(i)
+
+ @test Dates.firstdayofyear(dt) == first
+ @test Dates.firstdayofyear(DateTime(dt)) == DateTime(first)
+
+ @test Dates.lastdayofyear(dt) == last
+ @test Dates.lastdayofyear(DateTime(dt)) == DateTime(last)
+end
 
 # Adjusters
 # Adjuster Constructors

test/dates/io.jl

@@ -176,6 +176,17 @@ gg = "Jan-1996-15"
 f = "uuu-yyyy-dd"
 @test Dates.DateTime(gg,f) == dt
 @test Dates.format(dt,f) == gg
+hh = "1996#1#15"
+f = "yyyy#m#d"
+@test Dates.DateTime(hh,f) == dt
+@test Dates.format(dt,f) == hh
+
+# test prefix.
+s = "/1996/1/15"
+f = "/yyyy/m/d"
+@test Dates.DateTime(s,f) == dt
+@test Dates.format(dt,f) == s
+@test Dates.DateTime("1996/1/15",f) == dt
 
 # from Jiahao
 @test Dates.Date("2009年12月01日","yyyy年mm月dd日") == Dates.Date(2009,12,1)