[BEAM-10514] Make sure Interactive Beam cache file path length does not exceed OS limits

• Type: Improvement
• Status: Resolved
• Priority: P3
• Resolution: Fixed
• Fix Version/s: Missing
• Component/s: runner-py-interactive
• Assignee: Unassigned
• Reporter: Ning
• Created: 16/Jul/20 18:02
• Updated: 13/Jan/22 21:49
• Resolved: 05/Apr/21 17:05

Description

The path length limit in Linux is 4096.
The limit in windows API is 260.
If long file name support is enabled in windows, the limit will be 32767.

An example path that doesn't work on windows is c:\windows\temp\interactive-temp-xqj_fv471021776\cache-20-07-16-08_28_57\full\beam-temp-anonymous_pcollection_433934912-433934912-433938272-471021776-65e3a91ec73e11ea91b7e69191fc0bf0\284e0d91-ca8a-463c-a894-bde70cdec599.anonymous_pcollection_433934912-433934912-433938272-471021776 (length: 281).

Consider using obfuscation when mapping in-memory PCollections into files.
If long file name is needed, consider storing a reverse indexed mapping in memory.

Issue Links

• is duplicated by: BEAM-10627 tests fails on windows - interactive tests fails due to FileNotFoundError
• links to: GitHub Pull Request #12283