Incorrect results using multiple buckets with Azure Data Lake Storage URI

[nameexhaustion]

Checks

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of Polars.

Reproducible example

print(pl.scan_parquet("abfss://[email protected]/file1.parquet").explain())
# Parquet SCAN [abfss://account.dfs.core.windows.net/file1.parquet]
print(pl.scan_parquet("abfss://[email protected]/...").explain())
# Parquet SCAN [abfss://account.dfs.core.windows.net/file1.parquet]

Where file1.parquet exists only in bucket1

Log output

No response

Issue description

When using an Azure Data Lake Storage URI¹, the first bucket that gets used becomes hardcoded into the polars object store cache. Subsequent scan operations will ignore the bucket specified in the URI and always uses the first bucket.

Expected behavior

The 2nd scan_parquet in the example should fail, as the file does not exist in bucket2, instead of incorrectly scanning from bucket1

The resolved path should also include the bucket@ segment:

print(pl.scan_parquet("abfss://[email protected]/file1.parquet").explain())
# Parquet SCAN [abfss://[email protected]/file1.parquet]
print(pl.scan_parquet("abfss://[email protected]/...").explain())
# Expected at least 1 source

Installed versions

1.17.1

Footnotes

1. https://learn.microsoft.com/en-us/azure/storage/blobs/data-lake-storage-introduction-abfs-uri ↩