LigaData-ltsentsiper opened a new pull request, #55684: URL: https://github.com/apache/arrow/pull/55684
### Rationale for this change With a SAS token, `MakeDataLakeServiceClient` builds the `DataLakeServiceClient` from the blob URL. The Azure SDK sends Data Lake REST calls to the URL it is given, so the hierarchical namespace check goes to the blob endpoint, gets a 400, and `AzureFileSystem` treats a hierarchical namespace account as flat: `Move()` is "not implemented" and `DeleteFile()` fails. ### What changes are included in this PR? `MakeDataLakeServiceClient` uses `AccountDfsUrl(account_name) + sas_token_` for the SAS credential kind, like every other credential kind. ### Are these changes tested? Manually, against a real storage account with hierarchical namespace: a small program calling `AzureFileSystem` directly (`CreateDir`, write, `Move()`, `GetFileInfo`, `DeleteFile`, results in the issue) with stock and patched 26.0.1 and an account SAS, and an application that writes Parquet files and renames them (2,000,000 rows) with an account SAS and a container SAS. No automated test: the existing Azure tests run against Azurite, which has no hierarchical namespace. A unit test in `TestAzureOptions` could build `AzureOptions` with a SAS token and check that `MakeDataLakeServiceClient()->GetUrl()` uses the `dfs` host; I have not written or run it, and can add it if wanted. ### Are there any user-facing changes? SAS token users on hierarchical namespace accounts get the hierarchical namespace code paths (`Move()`, `DeleteFile()`, `CreateDir`) instead of the flat namespace ones. * Closes: GH-55683 🤖 Generated with [Claude Code](https://claude.com/claude-code) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
