juliageo/cftime.jl

Julia library for decoding time units conforming to the Climate and Forecasting (CF) netCDF conventions.

cftime.jl's Issues

Handle more second subdivisions

Describe the bug

Some netcdf files produced by python's xarray have time units nanoseconds since <datetime>.
Those are not yet handled by NCDataset via CFTime (Alexander-Barth/NCDatasets.jl#192).
Supporting additional second subdivisions in timedecode() would be appreciated, such as microsecond and nanosecond. This would also be consistent with python's handling, see Alexander-Barth/NCDatasets.jl#181 (comment).

To Reproduce

julia> using CFTime

julia> timedecode(1e9, "nanoseconds since 2000-01-01 00:00:00.001", "proleptic_gregorian")
ERROR: unknown units nanoseconds
Stacktrace:
 [1] error(s::String)
   @ Base ./error.jl:35
 [2] timeunits(#unused#::Type{DateTimeProlepticGregorian}, units::String)
   @ CFTime ~/.julia/packages/CFTime/n09Um/src/CFTime.jl:446
 [3] timedecode(#unused#::Type{DateTimeProlepticGregorian}, data::Float64, units::String)
   @ CFTime ~/.julia/packages/CFTime/n09Um/src/CFTime.jl:494
 [4] timedecode(data::Float64, units::String, calendar::String; prefer_datetime::Bool)
   @ CFTime ~/.julia/packages/CFTime/n09Um/src/CFTime.jl:545
 [5] timedecode(data::Float64, units::String, calendar::String)
   @ CFTime ~/.julia/packages/CFTime/n09Um/src/CFTime.jl:543
 [6] top-level scope
   @ REPL[6]:1

Expected behavior

Same result as with seconds:

julia> timedecode(1, "seconds since 2000-01-01 00:00:00.001", "proleptic_gregorian")
2000-01-01T00:00:01.001

Environment

• operating system: Ubunutu 18.04
• Julia version: 1.6.7 and 1.8.3
• CFTime: 0.1.2 and 0.1.3 (master)

Use default "days since 1900-01-01 00:00:00" as units kwarg for timeencode/decode?

It would be good to not have to specify this when automating conversions.

Edit: I may misunderstand exactly how this works... The end goal is to have reciprocal conversion to/from vectors of time types to strings without having to specify anything besides the DateTime objects so that NCDatasets can load and save vars of AbstractCFDateTime without and special casing. This is required for load/save to be easy and seemless in GeoData.jl.

Parsing delta_t to DatePeriod?

I would like to get time periods in julia DatePeriod object like Month(1) from netcdf time dimension delta_t fields like "0000-01-00 00:00:00". This would give correct time ranges in GeoData.jl.

Are there tools here that would already do that, or get some of the way?

handling of year 0 changed in python cftime

Describe the bug

The handling of year 0 changed in python cftime:

10/1/2021: Version 1.5.1 released. Changed default behavior of proleptic_gregorian to has_year_zero=T 
(since it is allowed in ISO-8601 and CF does not specify the year zero convention for this calendar). 
Raise warning message when trying to create a calendar that is not supported by CF version 1.9 
(no years < 1 allowed for 'standard'/'gregorian' or 'julian' calendars). 

To Reproduce

using CFTime, Dates
DateTimeProlepticGregorian(1,1,1) - Day(1)
# output: DateTimeProlepticGregorian(-001-12-31T00:00:00)

All calendars are affected expect DateTimeStandard and DateTimeJulian

Expected behavior

output should be DateTimeProlepticGregorian(0,12,31) as in python:

In [3]: cftime.DatetimeProlepticGregorian(1, 1, 1) + timedelta(days=-1)
Out[3]: cftime.DatetimeProlepticGregorian(0, 12, 31, 0, 0, 0, 0, has_year_zero=True)

Environment

• operating system: Ubuntu 20.04
• Julia version: julia 1.7.2, official binaries from https://julialang.org/downloads/

Full output

No additional output. All reference values from python and cftime 1.6.0:

In [3]: cftime.DatetimeProlepticGregorian(1, 1, 1) + timedelta(days=-1)
Out[3]: cftime.DatetimeProlepticGregorian(0, 12, 31, 0, 0, 0, 0, has_year_zero=True)

In [4]: cftime.DatetimeJulian(1, 1, 1) + timedelta(days=-1)
<ipython-input-4-e85e1427bb46>:1: CFWarning: this date/calendar/year zero convention is not supported by CF
  cftime.DatetimeJulian(1, 1, 1) + timedelta(days=-1)
Out[4]: cftime.DatetimeJulian(-1, 12, 31, 0, 0, 0, 0, has_year_zero=False)

In [5]: cftime.DatetimeGregorian(1, 1, 1) + timedelta(days=-1)
<ipython-input-5-d72578c40af5>:1: CFWarning: this date/calendar/year zero convention is not supported by CF
  cftime.DatetimeGregorian(1, 1, 1) + timedelta(days=-1)
Out[5]: cftime.DatetimeGregorian(-1, 12, 31, 0, 0, 0, 0, has_year_zero=False)

In [6]: cftime.DatetimeAllLeap(1, 1, 1) + timedelta(days=-1)
Out[6]: cftime.DatetimeAllLeap(0, 12, 31, 0, 0, 0, 0, has_year_zero=True)

In [7]: cftime.DatetimeNoLeap(1, 1, 1) + timedelta(days=-1)
Out[7]: cftime.DatetimeNoLeap(0, 12, 31, 0, 0, 0, 0, has_year_zero=True)

In [8]: cftime.Datetime360Day(1, 1, 1) + timedelta(days=-1)
Out[8]: cftime.Datetime360Day(0, 12, 30, 0, 0, 0, 0, has_year_zero=True)

daysinmonth(DateTimeNoLeap, 2) gives error

Running the function daysinmonth of month 2 gives error:

julia> daysinmonth(DateTimeNoLeap, 2)
ERROR: MethodError: no method matching rem(::Type{DateTimeNoLeap}, ::Int64)
Closest candidates are:
  rem(::Any, ::Any, ::RoundingMode{:ToZero}) at div.jl:67
  rem(::Any, ::Any, ::RoundingMode{:Down}) at div.jl:68
  rem(::Any, ::Any, ::RoundingMode{:Up}) at div.jl:69
  ...
Stacktrace:
 [1] isleapyear(y::Type)
   @ Dates /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.6/Dates/src/types.jl:159
 [2] daysinmonth(y::Type, m::Int64)
   @ Dates /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.6/Dates/src/types.jl:163
 [3] top-level scope
   @ REPL[12]:1

But other months work fine

julia> daysinmonth(DateTimeNoLeap, 1)
31

julia> daysinmonth(DateTimeNoLeap, 3)
31

julia> daysinmonth(DateTimeNoLeap, 4)
30

julia> daysinmonth(DateTimeNoLeap, 5)
31

julia> daysinmonth(DateTimeNoLeap, 6)
30

julia> daysinmonth(DateTimeNoLeap, 7)
31

julia> daysinmonth(DateTimeNoLeap, 8)
31

julia> daysinmonth(DateTimeNoLeap, 9)
30

julia> daysinmonth(DateTimeNoLeap, 10)
31

julia> daysinmonth(DateTimeNoLeap, 11)
30

I am using CFTime v0.1.1 of Julia version 1.6.1 on CentOS stream 8.

Request for "NO_CALENDAR" option for CFTime

Describe the bug

In GFDL model output, if absolute time is not important, a very commonly used option is to select the time option to be "NO_CALENDAR", which therefore has the following options as seen in the below example when using NCDatasets:

time   (360)
    Datatype:    Float64
    Dimensions:  time
    Attributes:
     long_name            = time
     units                = days since 0000-00-00 00:00:00
     cartesian_axis       = T
     calendar_type        = NO_CALENDAR
     calendar             = NO_CALENDAR
     bounds               = time_bounds

I cannot use NCDatasets.jl to read the time variable, because then it gives me:

ERROR: Unsupported calendar: no_calendar
Stacktrace:
 [1] error(::String) at ./error.jl:33
 [2] timetype(::String) at /Users/natgeo-wong/.julia/packages/CFTime/T83bb/src/CFTime.jl:594
 [3] timeunits(::String, ::String) at /Users/natgeo-wong/.julia/packages/CFTime/T83bb/src/CFTime.jl:610
 [4] getindex(::NCDataset, ::String) at /Users/natgeo-wong/.julia/packages/NCDatasets/Tb4eK/src/cfvariable.jl:281
 [5] top-level scope at none:0

Request

Is it possible to add a "NO_CALENDAR" option/struct to CFTime to accommodate model output?

parsing string to CF datetypes

I need to parse "21001231" into Date with the calender type of 365_day (DateTimeNoLeap).
Any idea how to do that?

Thank u.

DateTimeProlepticGregorian leap year in 300

Describe the bug

The year 300, which is not a leap year, has some odd behavior in the proleptic Gregorian calendar. Instead of going from February 28 to March 1, it goes from February 28 to February 29 to March 2, so March 1 is missing and replaced by a non existing leap day.

To Reproduce

using Dates, CFTime

isleapyear(300)  # false
isleapyear(304)  # true

daysinmonth(DateTimeProlepticGregorian, 300, 2)  # 28

Date(300, 2, 29)  # error: cannot create non existing day
g = DateTimeProlepticGregorian(300, 2, 29)  # can create in proleptic Gregorian
g = DateTimeProlepticGregorian(300, 3, 1)  # this creates the same day, February 29
g + Day(1)  # next day is March 2

DateTimeProlepticGregorian(2022, 1, 32)  # note that this creates February 1, not sure if desired

Expected behavior

I expect DateTimeProlepticGregorian(300, 2, 29) to error, or if desired create March 1.

Environment

CFTime v0.1.1

• operating system: Windows
• Julia version: e.g. julia 1.8.0-beta3, official binaries from https://julialang.org/downloads/]

Plotting with CFTime objects?

Sorry for not providing a fully reproducible example. I'm new to Julia.

When I call

using PyPlot
plot(time, timeseries)

where time is a Array{DateTimeNoLeap,1} object, I get this error:

Cannot convert DateTimeNoLeap to series data for plotting

Stacktrace:
 [1] error(::String) at ./error.jl:33
 [2] prepareSeriesData(::DateTimeNoLeap) at /Users/rpa/.julia/packages/Plots/2KhB2/src/series.jl:14
 [3] convertToAnyVector(::DateTimeNoLeap, ::Dict{Symbol,Any}) at /Users/rpa/.julia/packages/Plots/2KhB2/src/series.jl:26
 [4] (::Plots.var"#151#154"{Dict{Symbol,Any}})(::DateTimeNoLeap) at ./none:0
 [5] iterate(::Base.Generator{Array{DateTimeNoLeap,1},Plots.var"#151#154"{Dict{Symbol,Any}}}) at ./generator.jl:47
 [6] convertToAnyVector(::Array{DateTimeNoLeap,1}, ::Dict{Symbol,Any}) at /Users/rpa/.julia/packages/Plots/2KhB2/src/series.jl:41
 [7] macro expansion at /Users/rpa/.julia/packages/Plots/2KhB2/src/series.jl:129 [inlined]
 [8] apply_recipe(::Dict{Symbol,Any}, ::Type{Plots.SliceIt}, ::Array{DateTimeNoLeap,1}, ::Array{Float32,1}, ::Nothing) at /Users/rpa/.julia/packages/RecipesBase/zBoFG/src/RecipesBase.jl:275
 [9] _process_userrecipes(::Plots.Plot{Plots.GRBackend}, ::Dict{Symbol,Any}, ::Tuple{Array{DateTimeNoLeap,1},Array{Float32,1}}) at /Users/rpa/.julia/packages/Plots/2KhB2/src/pipeline.jl:83
 [10] _plot!(::Plots.Plot{Plots.GRBackend}, ::Dict{Symbol,Any}, ::Tuple{Array{DateTimeNoLeap,1},Array{Float32,1}}) at /Users/rpa/.julia/packages/Plots/2KhB2/src/plot.jl:178
 [11] #plot#137(::Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{(),Tuple{}}}, ::typeof(plot), ::Array{DateTimeNoLeap,1}, ::Vararg{Any,N} where N) at /Users/rpa/.julia/packages/Plots/2KhB2/src/plot.jl:57
 [12] plot(::Array{DateTimeNoLeap,1}, ::Array{Float32,1}) at /Users/rpa/.julia/packages/Plots/2KhB2/src/plot.jl:51
 [13] top-level scope at In[120]:3

Is there a recommended way to plot with CFTime object?

TagBot trigger issue

This issue is used to trigger TagBot; feel free to unsubscribe.

If you haven't already, you should update your TagBot.yml to include issue comment triggers.
Please see this post on Discourse for instructions and more details.

If you'd like for me to do this for you, comment TagBot fix on this issue.
I'll open a PR within a few hours, please be patient!

Use types for calendar arg in timedecode and timencode, and allow types in timetype

Requiring strings for the calendar type means the calling code has to convert the type to string if it want to encode the same calendar type.

I think adding timetype(dt::Type{<:AbstractCFDateTime}) = dt would acheive this?