Restore is not rolling back schema changes

[thomasfrederikhoeck]

Environment

Delta-rs version: 0.25.4

Binding: pythom

Environment:

• Cloud provider:
• OS: widnows
• Other:

---

Bug

What happened:
When restoring to a previous version schema changes are not rolled back.

What you expected to happen:
I excpet they are rolled back similar to PySpark:

How to reproduce it:

from deltalake import DeltaTable, Field, Schema

import polars as pl

df = pl.DataFrame({"colA":[1,2,3]})
DeltaTable.create("delta-table",schema=Schema([Field("colA","integer")]))
dt =  DeltaTable("delta-table")

df.write_delta("delta-table", mode="append")
dt =  DeltaTable("delta-table")
print(dt.schema())
print(dt.version())
dt.alter.add_columns(Field("colB","integer"))
print(dt.schema())
print(dt.version())
dt.restore(0)
print(dt.schema())
dt.to_pandas()
print(dt.to_pandas())

gives:

Schema([Field(colA, PrimitiveType("integer"), nullable=True)])
1
Schema([Field(colA, PrimitiveType("integer"), nullable=True), Field(colB, PrimitiveType("integer"), nullable=True)])
2
Schema([Field(colA, PrimitiveType("integer"), nullable=True), Field(colB, PrimitiveType("integer"), nullable=True)])
Empty DataFrame
Columns: [colA, colB]
Index: []

More details:

[ion-elgreco]

There is a PR addressing this: #3303

[aldder]

Hi @ion-elgreco

I'd like to reopen this issue since I'm still encountering the same error, even though the issue was previously closed.
I'm using 0.25.5 but I have the issue also with previous versions.

I'm seeing the same behavior as described in the original report, and the author’s minimal reproducible example still triggers the issue on my end.

Thanks!

[ion-elgreco]

@aldder the fix is not yet released