From 14eb39543a14d64d5df0534dd0ba0d63a2286b4d Mon Sep 17 00:00:00 2001
From: Henry Addison <henryaddison@users.noreply.github.com>
Date: Mon, 15 Nov 2021 12:05:34 +0000
Subject: [PATCH] open_mfdataset not playing well with encoding time when
 re-saving

annoying issue with xarray https://github.com/pydata/xarray/issues/2436
---
 deterministic/preprocessing/split_by_year.py | 19 ++++++++++++++-----
 deterministic/split-gcm.py                   |  6 +++---
 2 files changed, 17 insertions(+), 8 deletions(-)

diff --git a/deterministic/preprocessing/split_by_year.py b/deterministic/preprocessing/split_by_year.py
index d17c37e9..e0fbb731 100644
--- a/deterministic/preprocessing/split_by_year.py
+++ b/deterministic/preprocessing/split_by_year.py
@@ -2,26 +2,35 @@
 
 import cftime
 import xarray
+import glob
 
 """
 Split up an nc multifile dataset into single file per year.
 """
 class SplitByYear:
 
-    def __init__(self, input_dir, output_filepath_prefix, years = itertools.chain(range(1980, 2000), range(2020, 2040), range(2060, 2080))) -> None:
-        self.input_dir = input_dir
+    def __init__(self, input_filepath_prefix, output_filepath_prefix, years = itertools.chain(range(1980, 2000), range(2020, 2040), range(2060, 2080))) -> None:
+        self.input_filepath_prefix = input_filepath_prefix
         self.output_filepath_prefix = output_filepath_prefix
         self.years = years
 
         pass
 
-    def run(self):
-        output_files = []
+    def gcm_file(self, year):
+        if (year % 10) <= 8:
+            start = (year // 10) * 10 - 1
+        else:
+            start = year
+
+        end = start + 10
 
-        input = xarray.open_mfdataset(str(self.input_dir/"*.nc"))
+        return f"{start}1201-{end}1130.nc"
 
+    def run(self):
+        output_files = []
 
         for year in self.years:
+            input = xarray.load_dataset(f"{self.input_filepath_prefix}_{self.gcm_file(year)}.nc")
             single_year_input = input.sel(time=slice(cftime.Datetime360Day(year, 12, 1, 12, 0, 0, 0) , cftime.Datetime360Day(year+1, 11, 30, 12, 0, 0, 0)))
 
             output_filepath = f"{self.output_filepath_prefix}_{year}1201-{year+1}1130.nc"
diff --git a/deterministic/split-gcm.py b/deterministic/split-gcm.py
index adbb236b..f2396c0b 100644
--- a/deterministic/split-gcm.py
+++ b/deterministic/split-gcm.py
@@ -7,8 +7,8 @@
 def get_args():
     parser = argparse.ArgumentParser(description='Regrid GCM data to match the CPM data',
                                      formatter_class=argparse.ArgumentDefaultsHelpFormatter)
-    parser.add_argument('--input', dest='input_dir', type=Path, required=True,
-                        help='Path to directory holding raw files')
+    parser.add_argument('--input-prefix', dest='input_prefix', type=Path, required=True,
+                        help='Prefix of input files to split up (so filepath up to the date part)')
     parser.add_argument('--output-prefix', dest='output_prefix', type=str, required=True,
                         help='Prefix of output files including directory path')
     parser.add_argument('--years', dest='years', nargs='+', type=int, required=True,
@@ -22,6 +22,6 @@ def get_args():
 
     os.makedirs(os.path.dirname(args.output_prefix), exist_ok=True)
 
-    outputs = SplitByYear(args.input_dir, args.output_prefix, args.years).run()
+    outputs = SplitByYear(args.input_prefix, args.output_prefix, args.years).run()
 
     print(outputs)