Unzipping and filtering OpenURLWithFSSpec ouput before OpenWithXarray

[Riccardo7-DS]

I am trying to retreive data from the Landsaf portal for MSG SEVIRI radiometer. https://datalsasaf.lsasvcs.ipma.pt/PRODUCTS/EPS/ENDVI10/ENVI/ However, the data is currently zipped and contains multiple files of different formats. Using a similar approach of #659 I am trying to open the unzipped and filtered file with the open_rasterio kwarg within OpenWithXarray class.

the error I currently get is: TypeError: cannot unpack non-iterable NoneType object for line 98 of ".../pangeo_forge_recipes/transforms.py" because the OpenWithXarray class fails to open the output of the custom class UnzipFilter I have created. Any hints on how to solve this as I am still a beginner with pangeo_forge_recipes?

from pangeo_forge_recipes.patterns import ConcatDim, FilePattern
from pangeo_forge_recipes.transforms import PrepareZarrTarget, OpenURLWithFSSpec, OpenWithXarray, StoreToZarr

from pangeo_forge_recipes.transforms import _add_keys, MapWithConcurrencyLimit
from pangeo_forge_recipes.openers import OpenFileType
import pangeo_forge_recipes

import apache_beam as beam
import zipfile
import os
import pandas as pd
import fsspec
from typing import Union, Optional, Tuple
import aiohttp
import io
from dataclasses import dataclass, field

@dataclass
class UnzipFilter(beam.PTransform):
    
    num: Optional[int] = 1
    file_format: Optional[str] = None
    file_name: Optional[str] = None
    file_substring: Optional[str] = None

    def expand(self, pcoll):
        refs = pcoll  | "Unzip and filter" >> beam.Map(
            _unzip_and_filter,
            num=self.num,
            file_format=self.file_format,
            file_name=self.file_name,
            file_substring=self.file_name,
        )
        return refs

def _unzip_and_filter(
        response: Tuple[pangeo_forge_recipes.types.Index, OpenFileType], 
        num:int=1, 
        file_format:Union[None,str]=None,
        file_name:Union[None,str]=None, 
        file_substring:Union[None,str]=None):
    
    import io
    with response[1] as f:
        zip_contents = f.read()

    # Step 2: Create a BytesIO object to treat the contents as an in-memory file
    zip_buffer = io.BytesIO(zip_contents)

    # Step 3: Use zipfile to extract the files from the in-memory buffer
    with zipfile.ZipFile(zip_buffer, 'r') as zip_ref:
        # Filter files based on the specified pattern
        if file_name is not None:
            zip_file_list = [file for file in zip_ref.namelist() 
                             if file_name == file]
        elif file_substring is not None:
            zip_file_list = [file for file in zip_ref.namelist() 
                             if file_substring in file]
        elif file_format is not None:
            zip_file_list = [file for file in zip_ref.namelist() 
                             if file.endswith(file_format)]

        if num ==1:
            zip_ref.read(zip_file_list[0]) 
        else:
            raise NotImplementedError

def make_url(time):
    url = URL_FORMAT.format(time=time, day=time.day, 
                             month=time.month, 
                             year=time.year)
    return url

start_date = '2008-01-01'
end_date = '2008-04-21'

# Create a date range for the 1st, 11th, and 21st of each month
date_range = pd.date_range(start=start_date, end=end_date, freq='10D')
# Filter the dates to include only the 1st, 11th, and 21st
dates = date_range[date_range.day.isin([1, 11, 21])]

username = config["LANDSAF"]["user"]
password = config["LANDSAF"]["password"]

path = config["DEFAULT"]["output"]
target_store = "output_file.zarr"

URL_FORMAT = (
    "https://datalsasaf.lsasvcs.ipma.pt/PRODUCTS/EPS/ENDVI10/ENVI/{year:4d}/{month:02d}/{day:02d}/METOP_AVHRR_{time:%Y%m%d}_S10_AFR_V200.zip"
)
time_concat_dim = ConcatDim("time", dates, nitems_per_file=1)
pattern = FilePattern(make_url, time_concat_dim)
url = make_url(dates[0])

open_kwarg = {"auth" : aiohttp.BasicAuth(username, password), "num" : 10}


recipe = (
    beam.Create(pattern.items())
    | OpenURLWithFSSpec(open_kwargs=open_kwarg)
    | UnzipFilter(file_format=".tif")
    | OpenWithXarray(xarray_open_kwargs={'engine': 'rasterio'})
    | PrepareZarrTarget(os.path.join(path, target_store))
    | StoreToZarr(
        target_root=path,
        store_name=target_store,
        combine_dims=pattern.combine_dim_keys,
    )
)

from apache_beam.pipeline import PipelineOptions
with beam.Pipeline() as p:
    p | recipe

[thodson-usgs]

It's a bit hacky but have a look at my workaround: https://github.com/hytest-feedstocks/ssebop-feedstock

Note the extra zip sugar in the url pattern.

[Riccardo7-DS]

Thanks @thodson-usgs , it's a nice workaround