Skip to content

read_vrt chunks mode materializes the full mosaic before chunking #1798

@brendancol

Description

@brendancol

Describe the bug

read_vrt(chunks=...) returns a dask-backed DataArray, but it first eagerly assembles the full VRT into memory and only then chunks the result. This makes the API look lazy while still carrying full eager memory risk for large mosaics.

Expected behavior

Either read_vrt(chunks=...) should be truly lazy, building dask tasks over VRT windows, or the public contract should be changed to make the eager materialization explicit.

Screenshots

Not applicable.

Desktop

Not applicable.

Smartphone

Not applicable.

Additional context

This affects large VRT mosaics and can produce out-of-memory failures before dask has a chance to help.

Metadata

Metadata

Assignees

No one assigned

    Labels

    bugSomething isn't workingoomOut-of-memory risk with large datasetsperformancePR touches performance-sensitive code

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions