How to spatially concatenate (along latitude and longitude) using open_virtual_mfdataset

[PouriaRezz]

Hi Virtualizarr team,

I'm testing Virtualizarr for the first time and would like to virtually concatenate a large collection of NetCDF (.nc) files hosted remotely (over HTTP) without loading any data into memory. Here's what I'm doing:

import warnings
import fsspec
from virtualizarr.parsers import HDFParser
from virtualizarr import open_virtual_mfdataset
from obstore.store import HTTPStore

warnings.filterwarnings("ignore", category=UserWarning)

fs = fsspec.filesystem('http')
nc_files = sorted(fs.glob("https://sense4fire.eu/database/v02/S4F.CCILC_FireCCI51/Europe-reg_333m/bm_herb/*.nc"))
store = HTTPStore("https://sense4fire.eu")

vds = open_virtual_mfdataset(
    nc_files[:30],
    parser=HDFParser(),
    object_store=store,
    combine_attrs="override",
    data_vars="minimal",
    coords="minimal"
)

Unfortunately, I run into this error:

TypeError: Could not find a Chunk Manager which recognises type <class 'virtualizarr.manifests.array.ManifestArray'>

From the traceback, it seems like xarray tries to call .compute() or .load() internally during the combine_by_coords logic.
I also tried without the data_vars/coords arguments:

vds = open_virtual_mfdataset(
    nc_files[:30],
    parser=HDFParser(),
    object_store=store,
    combine_attrs="override"
)

But then i get:

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

My Understanding:

Virtualizarr is meant to allow lazy access without loading data. But xarray.combine_by_coords tries to compare variables internally using .compute(), which causes ManifestArray to break because it's not integrated with any known chunk manager

Question:

Is it possible to concatenate these files fully lazily, i.e., without triggering any .compute() or .load() on the underlying arrays?

Let me know if I'm missing something or if you have an idea how to do it right i would be happy to hear it