I am trying to filter out the URI column from a parquet file having over 50 million rows containing empty string using

import polars as pl

lf = pl.scan_parquet("data.parquet")
lf.filter(pl.col("URI") == "").collect()

Output:

shape: (0, 3)
┌─────┬────────┬───────────┐
│ URI ┆ REMARK ┆ TIMESTAMP │
│ --- ┆ ---    ┆ ---       │
│ str ┆ str    ┆ i64       │
╞═════╪════════╪═══════════╡
└─────┴────────┴───────────┘

Luckily I had labelled the rows with empty string URI in column REMARK with NO URI so,

lf.filter(pl.col("REMARK") == "NO URI").collect()

yields:

shape: (7_767, 3)
┌─────┬────────┬────────────┐
│ URI ┆ REMARK ┆ TIMESTAMP  │
│ --- ┆ ---    ┆ ---        │
│ str ┆ str    ┆ i64        │
╞═════╪════════╪════════════╡
│     ┆ NO URI ┆ 1759257000 │
│     ┆ NO URI ┆ 1759257000 │
│     ┆ NO URI ┆ 1759257000 │
│     ┆ NO URI ┆ 1759257000 │
│     ┆ NO URI ┆ 1759257000 │
│ …   ┆ …      ┆ …          │
│     ┆ NO URI ┆ 1759257000 │
│     ┆ NO URI ┆ 1759257000 │
│     ┆ NO URI ┆ 1759257000 │
│     ┆ NO URI ┆ 1759257000 │
│     ┆ NO URI ┆ 1759257000 │
└─────┴────────┴────────────┘

Also for confirmation that the URI column string is just empty string

len(lf.filter(pl.col("REMARK") == "NO URI").collect()["URI"][0]) # Outputs 0

Is this is a bug in polars or have I missed some important info, and how do I get the rows with empty string?

Python version: 3.14.2
Polars version: 1.35.2