to_zarr: region not recognised as dataset dimensions

[Boorhin]

What happened:
I am trying to write into a zarr a dataset by reading each timesteps and write this time step into a zarr store. For that I prepared the dataset and filled the variable values with 0s and wrote the store with mode='a'. I then perform the operations I need on each variables and try to write the time step of the dataset.

ds.isel(time=t).to_zarr(outfile, region={"time": t})

However I received this error message:
ValueError: all keys in ``region`` are not in Dataset dimensions, got ['time'] and ['cell', 'face', 'layer', 'max_cell_node', 'max_face_nodes', 'node', 'siglay']
But

In: ds.dims
Out: Frozen(SortedKeysDict({'time': 1465, 'node': 112015, 'layer': 6, 'face': 198364, 'max_face_nodes': 3, 'cell': 991820, 'max_cell_node': 6, 'siglay': 6}))

Checking in the API, it comes from

~/.local/lib/python3.8/site-packages/xarray/backends/api.py in _validate_append_dim_and_encoding(ds_to_append, store, append_dim, region, encoding, **open_kwargs)
   1330         for k, v in region.items():
   1331             if k not in ds_to_append.dims:
-> 1332                 raise ValueError(
   1333                     f"all keys in ``region`` are not in Dataset dimensions, got "
   1334                     f"{list(region)} and {list(ds_to_append.dims)}"

What you expected to happen:
Incremently append data to the zarr store

Minimal Complete Verifiable Example:

import xarray as xr
from datetime import datetime,timedelta
import numpy as np
dt= datetime.now()
times= np.arange(dt,dt+timedelta(days=6), timedelta(hours=1))
nodesx,nodesy,layers=np.arange(10,50), np.arange(10,50)+15, np.arange(10)
ds=xr.Dataset()
ds.coords['time']=('time', times)
ds.coords['node_x']=('node', nodesx)
ds.coords['node_y']=('node', nodesy)
ds.coords['layer']=('layer', layers)
outfile='my_zarr'
varnames=['potato','banana', 'apple']
for var in varnames:
    ds[var]=(('time', 'layer', 'node'), np.zeros((len(times), len(layers),len(nodesx))))
ds.to_zarr(outfile, mode='a')
for t in range(len(times)):
    for var in varnames:
    	ds[var].isel(time=t).values += np.random.random((len(layers),len(nodesx)))
    ds.isel(time=t).to_zarr(outfile, region={"time": t})

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-25-8baf03aa01a3> in <module>
      2     for var in varnames:
      3         ds[var].isel(time=t).values += np.random.random((len(layers),len(nodesx)))
----> 4     ds.isel(time=t).to_zarr(outfile, region={"time": t})
      5 

~/.local/lib/python3.8/site-packages/xarray/core/dataset.py in to_zarr(self, store, chunk_store, mode, synchronizer, group, encoding, compute, consolidated, append_dim, region)
   1743             encoding = {}
   1744 
-> 1745         return to_zarr(
   1746             self,
   1747             store=store,

~/.local/lib/python3.8/site-packages/xarray/backends/api.py in to_zarr(dataset, store, chunk_store, mode, synchronizer, group, encoding, compute, consolidated, append_dim, region)
   1457     if mode == "a":
   1458         _validate_datatypes_for_zarr_append(dataset)
-> 1459         _validate_append_dim_and_encoding(
   1460             dataset,
   1461             store,

~/.local/lib/python3.8/site-packages/xarray/backends/api.py in _validate_append_dim_and_encoding(ds_to_append, store, append_dim, region, encoding, **open_kwargs)
   1330         for k, v in region.items():
   1331             if k not in ds_to_append.dims:
-> 1332                 raise ValueError(
   1333                     f"all keys in ``region`` are not in Dataset dimensions, got "
   1334                     f"{list(region)} and {list(ds_to_append.dims)}"

ValueError: all keys in ``region`` are not in Dataset dimensions, got ['time'] and ['layer', 'node']

Anything else we need to know?:

Environment:

Output of xr.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.8.10 (default, Sep 28 2021, 16:10:42)
[GCC 9.3.0]
python-bits: 64
OS: Linux
OS-release: 5.4.0-91-generic
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_GB.UTF-8
LOCALE: en_GB.UTF-8
libhdf5: 1.10.4
libnetcdf: 4.7.3

xarray: 0.16.2
pandas: 1.2.2
numpy: 1.17.4
scipy: 1.6.2
netCDF4: 1.5.3
pydap: None
h5netcdf: None
h5py: None
Nio: None
zarr: 2.10.2
cftime: 1.1.0
nc_time_axis: 1.2.0
PseudoNetCDF: None
rasterio: None
cfgrib: None
iris: None
bottleneck: None
dask: 2021.11.2
distributed: 2021.11.2
matplotlib: 3.1.2
cartopy: 0.18.0
seaborn: None
numbagg: None
pint: None
setuptools: 45.2.0
pip: 20.0.2
conda: None
pytest: 6.2.1
IPython: 7.13.0
sphinx: None

[observingClouds]

Hi @Boorhin,
I just ran into the same issue. The region argument has to be of type slice, in your case slice(t) instead of just t works:

import xarray as xr
from datetime import datetime,timedelta
import numpy as np
dt= datetime.now()
times= np.arange(dt,dt+timedelta(days=6), timedelta(hours=1))
nodesx,nodesy,layers=np.arange(10,50), np.arange(10,50)+15, np.arange(10)
ds=xr.Dataset()
ds.coords['time']=('time', times)
ds.coords['node_x']=('node', nodesx)
ds.coords['node_y']=('node', nodesy)
ds.coords['layer']=('layer', layers)
outfile='my_zarr'
varnames=['potato','banana', 'apple']
for var in varnames:
    ds[var]=(('time', 'layer', 'node'), np.zeros((len(times), len(layers),len(nodesx))))
ds.to_zarr(outfile, mode='a')
for t in range(len(times)):
    for var in varnames:
    	ds[var].isel(time=slice(t)).values += np.random.random((len(layers),len(nodesx)))
    ds.isel(time=slice(t)).to_zarr(outfile, region={"time": slice(t)})

This leads however to another issue:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-52-bb3d2c1adc12> in <module>
     18     for var in varnames:
     19         ds[var].isel(time=slice(t)).values += np.random.random((len(layers),len(nodesx)))
---> 20     ds.isel(time=slice(t)).to_zarr(outfile, region={"time": slice(t)})

~/.local/lib/python3.8/site-packages/xarray/core/dataset.py in to_zarr(self, store, chunk_store, mode, synchronizer, group, encoding, compute, consolidated, append_dim, region, safe_chunks)
   2029             encoding = {}
   2030 
-> 2031         return to_zarr(
   2032             self,
   2033             store=store,

~/.local/lib/python3.8/site-packages/xarray/backends/api.py in to_zarr(dataset, store, chunk_store, mode, synchronizer, group, encoding, compute, consolidated, append_dim, region, safe_chunks)
   1359 
   1360     if region is not None:
-> 1361         _validate_region(dataset, region)
   1362         if append_dim is not None and append_dim in region:
   1363             raise ValueError(

~/.local/lib/python3.8/site-packages/xarray/backends/api.py in _validate_region(ds, region)
   1272     ]
   1273     if non_matching_vars:
-> 1274         raise ValueError(
   1275             f"when setting `region` explicitly in to_zarr(), all "
   1276             f"variables in the dataset to write must have at least "

ValueError: when setting `region` explicitly in to_zarr(), all variables in the dataset to write must have at least one dimension in common with the region's dimensions ['time'], but that is not the case for some variables here. To drop these variables from this dataset before exporting to zarr, write: .drop(['node_x', 'node_y', 'layer'])

Here, the solution is however provided with the error message. Following the instructions, the snippet below finally works (as far as I can tell):

import xarray as xr
from datetime import datetime,timedelta
import numpy as np
dt= datetime.now()
times= np.arange(dt,dt+timedelta(days=6), timedelta(hours=1))
nodesx,nodesy,layers=np.arange(10,50), np.arange(10,50)+15, np.arange(10)
ds=xr.Dataset()
ds.coords['time']=('time', times)
# ds.coords['node_x']=('node', nodesx)
# ds.coords['node_y']=('node', nodesy)
# ds.coords['layer']=('layer', layers)
outfile='my_zarr'
varnames=['potato','banana', 'apple']
for var in varnames:
    ds[var]=(('time', 'layer', 'node'), np.zeros((len(times), len(layers),len(nodesx))))
ds.to_zarr(outfile, mode='a')
for t in range(len(times)):
    for var in varnames:
    	ds[var].isel(time=slice(t)).values += np.random.random((len(layers),len(nodesx)))
    ds.isel(time=slice(t)).to_zarr(outfile, region={"time": slice(t)})

Maybe one would like to generalise region in api.py to allow for single indices or throw a hint in case an a type different to a slice is provided.

Cheers

[Boorhin]

I don't get the second crash. It is not true that these variables are not in common, they are the coordinates of each of the variables. They are all made the same. This is a typical example of an unstructured grid backup. Meanwhile I found an alternate solution which is also better for memory management. I think the documentation example doesn't actually work.
I will try to formulate my trick but that's not using this particular method of region that is not functioning as it should in my opinion.

[observingClouds]

You are right, the coordinates should not be dropped.

I think the function _validate_region has a bug. Currently it checks for all ds.variables if at least one of their dimensions agrees with the ones given in the region argument. However, ds.variables also returns the coordinates. However, we actually only want to check if the ds.data_vars have a dimension intersecting with the given region.

Changing the function to

def _validate_region(ds, region):
    if not isinstance(region, dict):
        raise TypeError(f"``region`` must be a dict, got {type(region)}")

    for k, v in region.items():
        if k not in ds.dims:
            raise ValueError(
                f"all keys in ``region`` are not in Dataset dimensions, got "
                f"{list(region)} and {list(ds.dims)}"
            )
        if not isinstance(v, slice):
            raise TypeError(
                "all values in ``region`` must be slice objects, got "
                f"region={region}"
            )
        if v.step not in {1, None}:
            raise ValueError(
                "step on all slices in ``region`` must be 1 or None, got "
                f"region={region}"
            )

    non_matching_vars = [
        k for k, v in ds.data_vars.items() if not set(region).intersection(v.dims)
    ]
    if non_matching_vars:
        raise ValueError(
            f"when setting `region` explicitly in to_zarr(), all "
            f"variables in the dataset to write must have at least "
            f"one dimension in common with the region's dimensions "
            f"{list(region.keys())}, but that is not "
            f"the case for some variables here. To drop these variables "
            f"from this dataset before exporting to zarr, write: "
            f".drop({non_matching_vars!r})"
        )

seems to work.

[shoyer]

The reason why this isn't allowed is because it's ambiguous what to do with the other variables that are not restricted to the region (['cell', 'face', 'layer', 'max_cell_node', 'max_face_nodes', 'node', 'siglay'] in this case).

I can imagine quite a few different ways this behavior could be implemented:

1. Ignore these variables entirely.
2. Ignore variables if they also already exist, but write new ones.
3. Write or overwrite both new and existing these variables.
4. Write new variables. Ignore existing variables only if they already exist with the same values, and if not, raise an error.

I believe your proposal here (removing these checks from _validate_region) would achieve (3), but I'm not sure that's the best option.

(4) seems like perhaps the most user-friendly option, but checking existing variables can add significant overhead. When experimenting adding region support Xarray-Beam, I found many cases where it was easy to inadvertently make large parallel pipelines much slower by downloaded existing variables.

The current solution is not to do any of these, and to force the user to make an explicit choice by dropping new variables, or write them in a separate call to to_zarr. I think it would also be OK to let a user explicitly opt-in to one of these behaviors, but I don't think guessing what the user wants would be ideal.

[Boorhin]

If Xarray/zarr is to replace netcdf, appending by time step is really an important feature
Most (all?) numerical models will output results per time step onto a multidimensional grid with different variables
Said grid will also have other parameters that will help rebuild the geometry or follow standards, like CF and Ugrid (The things that you are supposed to drop). The geometry of the grid is computed at the initialisation of the model. It is a bit counter intuitive to get rid of it for incremental backups especially that each write will not concern this part of the file.
What I do at the moment is that I create a first dataset at the final dimension based on dummy dask arrays
Export it to_zarr with compute = False

With a buffer system, I create a new dataset for each buffer with the right data at the right place meaning only the time interval concerned and I write to_zarr with the region attribute
I flush the buffer dataset after it being written.

At the end I write all the parameters before closing the main dataset.

To my knowledge, that's the only method which works.

[d70-t]

I'm trying to picture some usage scenarios based on incrementally adding timesteps to data on store. I hope these might help to answer questions from above. In particular, I think that append and region options of to_zarr will imply different usage patterns, so might lead to different answers, and mixing terms might lead to confusion.

I'll use the following dataset for demonstration code:

ds = xr.Dataset({
    "T": (("time", "x"), [[1.,2.,3.],[11.,12.,13.]]),
}, coords={
    "time": (("time",), [21., 22.]),
    "x": (("x",), [100., 200., 300.])
}).chunk({"time": 1})

append

The purpose of append is to add (one or many) elements along one dimension after the end of all currently existing elements. This implies a read-modify-write cycle to at least the total shape of the array. Furthermore, the place to write new chunks is determined by the current shape of the existing array. Due to these implications, it doesn't seem to be useful to try append in parallel (it would become ambiguous where to write) and it doesn't seem to be too useful (but possible) to only write some of the variables defined on the append-dimension, because all other variables would implicitly be filled with fill_value and those places couldn't be filled with another append anymore.

As a consquence, append-mode writes will always have to be sequential and writes to objects shared touched by multiple append calls will always have a defined behaviour, even if they are modified / overwritten with each call. Creating and appending works as follows:

# writes 0-sized time-dimension, so only metadata and non-time dependent variables
ds.isel(time=slice(0,0)).to_zarr("test_append.zarr")

!tree -a test_append.zarr

ds.isel(time=slice(0,1)).to_zarr("test_append.zarr", mode="a", append_dim="time")
ds.isel(time=slice(1,2)).to_zarr("test_append.zarr", mode="a", append_dim="time")

print()
print("final dataset:")
!tree -a test_append.zarr

Output

test_append.zarr
├── .zattrs
├── .zgroup
├── .zmetadata
├── T
│   ├── .zarray
│   └── .zattrs
├── time
│   ├── .zarray
│   └── .zattrs
└── x
    ├── .zarray
    ├── .zattrs
    └── 0

3 directories, 10 files

final dataset:
test_append.zarr
├── .zattrs
├── .zgroup
├── .zmetadata
├── T
│   ├── .zarray
│   ├── .zattrs
│   ├── 0.0
│   └── 1.0
├── time
│   ├── .zarray
│   ├── .zattrs
│   ├── 0
│   └── 1
└── x
    ├── .zarray
    ├── .zattrs
    └── 0

3 directories, 14 files

In this case, x would be overwritten with each append call, but the behaviour is well defined as we will only ever append sequentially, so whatever the last write writes into x will be the final result, e.g. [1, 2, 3] in the following case:

ds.isel(time=slice(0,1)).to_zarr("test_append.zarr", mode="a", append_dim="time")
ds2 = ds.assign(x=[1,2,3])
ds2.isel(time=slice(1,2)).to_zarr("test_append.zarr", mode="a", append_dim="time")

If instead, x shouldn't be overwritten, it's possible to append using:

ds.drop(["x"]).isel(time=slice(0,1)).to_zarr("test_append.zarr", mode="a", append_dim="time")
ds.drop(["x"]).isel(time=slice(1,2)).to_zarr("test_append.zarr", mode="a", append_dim="time")

This also works already with current xarray and has well defined behaviour. However, if there are many time-independent variables, it might be easier if something like .drop_if_not("time") or something similar would be available.

region

region behaves quite differently from append. It does not modify the shape of the arrays and it does not depend on the shape's value to determine where to write new data (it requires user input to do so). This generally enables parallel writes to the same dataset (if only distinct chunks are touched). But as metadata (e.g. shape) is still shared, updates to metadata must be done in a coordinated (likely sequential) manner.

Generally, the workflow with region would imply writing the metadata once and maybe update it from time to time but sequentially (e.g. on a coordinating node) and write all the chunks in parallel on worker nodes, while carefully ensuring that no common chunks are overwritten. Let's see how this might look like:

ds.to_zarr("test.zarr", compute=False, encoding={"time": {"chunks": [1]}})
!rm test.zarr/time/0
!rm test.zarr/time/1

!tree -a test.zarr

# NOTE: these may run in parallel (even if that's not useful in time, but region might also be in time and space)
ds.drop(['x']).isel(time=slice(0,1)).to_zarr("test.zarr", mode="r+", region={"time": slice(0,1)})
ds.drop(['x']).isel(time=slice(1,2)).to_zarr("test.zarr", mode="r+", region={"time": slice(1,2)})

print()
print("final dataset:")
!tree -a test.zarr

Output

test.zarr
├── .zattrs
├── .zgroup
├── .zmetadata
├── T
│   ├── .zarray
│   └── .zattrs
├── time
│   ├── .zarray
│   └── .zattrs
└── x
    ├── .zarray
    ├── .zattrs
    └── 0

3 directories, 10 files
final dataset:
test.zarr
├── .zattrs
├── .zgroup
├── .zmetadata
├── T
│   ├── .zarray
│   ├── .zattrs
│   ├── 0.0
│   └── 1.0
├── time
│   ├── .zarray
│   ├── .zattrs
│   ├── 0
│   └── 1
└── x
    ├── .zarray
    ├── .zattrs
    └── 0

3 directories, 14 files

The above works and as far as I understand does what we'd want for parallel writes. It also avoids the mentioned ambiguous cases (due to the drop(['x']) statements). However this case is even more cumbersome to write than in the append case. The parallel writes might benefit from again from something like .drop_if_not("time") (which probably can't be optional in this case due to ambiguity). But what's even more problematic is the initial write of array metadata. In order to start building the dataset, I'll have to scaffold an (potentially not yet computed) Dataset of full size and use compute=False to write only metadata. However, this fails for coordinate variables (like time), because those are eagerly loaded and will still be written out. That's why I've removed those chunks in the example above.

If region should be used for parallel append, then there must be some process on a coordinating node which updates the metadata keys (at least by increasing the shape). I don't yet see how that could be written nicely using xarray.

---

So based on these two kinds of tasks, it seems to me that the actual append and region write-modes of to_zarr are already doing what they should do, but there could be some more convenience functions which would make those tasks much simpler:

• some method like drop_if_not (maybe with a better name) which drops all the things we don't want to keep (maybe we should call it keep instead of drop). This method would essentially result in and simplify mode 1 in @shoyer's answer, which I'd argue is what we actually want in both use cases, becasue the dropped data would already have been written by the coordinating process. I'd believe that mode 1 shouldn't be the default for to_zarr though, because silently dropping data from being written isn't nice to the user.
• some better tooling for writing and updating zarr dataset metadata (I don't know if that would fit in the realm of xarray though, as it looks like handling Datasets without content. For "appending" metadata, I really don't know how I'd picture this propery in xarray world.)

[d70-t]

While testing a bit further, I found another case which might potentially be dangerous:

# ds is the same as above, but chunksize is {"time": 1, "x": 1}

# once on the coordinator
ds.to_zarr("test.zarr", compute=False, encoding={"time": {"chunks": [1]}, "x": {"chunks": [1]}})
# in parallel
ds.isel(time=slice(0,1), x=slice(0,1)).to_zarr("test.zarr", mode="r+", region={"time": slice(0,1), "x": slice(0,1)})
ds.isel(time=slice(0,1), x=slice(1,2)).to_zarr("test.zarr", mode="r+", region={"time": slice(0,1), "x": slice(1,2)})
ds.isel(time=slice(0,1), x=slice(2,3)).to_zarr("test.zarr", mode="r+", region={"time": slice(0,1), "x": slice(2,3)})
ds.isel(time=slice(1,2), x=slice(0,1)).to_zarr("test.zarr", mode="r+", region={"time": slice(1,2), "x": slice(0,1)})
ds.isel(time=slice(1,2), x=slice(1,2)).to_zarr("test.zarr", mode="r+", region={"time": slice(1,2), "x": slice(1,2)})
ds.isel(time=slice(1,2), x=slice(2,3)).to_zarr("test.zarr", mode="r+", region={"time": slice(1,2), "x": slice(2,3)})

This example doesn't produce any error, but the time and x coordinates are re-written multiple times without any warning. However, I don't yet know how a proper error / warning should be generated in this case. Maybe the check must be if every written variable touches all region-ed dimensions? But maybe thats overly restrictive?

[Boorhin]

I have looked at these examples and I still don't manage to make it work in the real world.
I find append the most logical but I have attributes attached to a dataset that I don't seem to be able to drop before appending. This generates this error:
ValueError: failed to prevent overwriting existing key _FillValue in attrs. This is probably an encoding field used by xarray to describe how a variable is serialized. To proceed, remove this key from the variable's attributes manually.
However, I cannot find a way of getting rid of this attribute

[Boorhin]

I did make
ds.attrs={}
but at each appending I get a warning

/opt/conda/lib/python3.7/site-packages/xarray/core/dataset.py:2050: SerializationWarning: saving variable None with floating point data as an integer dtype without any _FillValue to use for NaNs
  safe_chunks=safe_chunks,

[d70-t]

I don't yet know a proper answer, but there'd be three observations I have:

• The ValueError seems to be related to the handling of CF-Conventions. I don't yet know if that's independent of this issue or if the error only appears in conjunction with this issue.
• As far as I understand, appending should be possible without dropping anything (while potentially overwriting some things).
• It shouldn't be possible to change _FillValue during appends, because that might require rewriting everything previously written, which you likely don't want. So if _FillValue is different on the append-call, I'd want xarray to produce an error.

[Boorhin]

The _FillValue is always the same (np.nan) and specified when I reproject with rioxarray.
so I don't understand the first error then.
The thing is that the _fillvalue is attached to a variable not the whole dataset. But it never change. Not too sure what to do

[d70-t]

🤷 can't help any further without a minimal reproducible example here...