What causes "input string does not contain DataFrame or Series" in pola-rs/polars?

Make sure the string is the verbatim output of print(df) / repr(df) or repr(s) for a pl.Series, including the border rows and the 'shape: (n, m)' header

pola-rs/polars · error · ValueError

input string does not contain DataFrame or Series

Error message

input string does not contain DataFrame or Series

What it means

Raised by polars.from_repr when the input string matches neither the DataFrame table layout (found via _find_df_repr / tbl_repr_type being None) nor the Series regex (shape + 'Series:' header). from_repr reconstructs data solely by parsing the printed repr, so arbitrary text, partial copies, or differently formatted output fail both parsers and hit the terminal ValueError.

Source

Thrown at py-polars/src/polars/convert/general.py:896

    ... )
    >>> s.to_list()
    [True, False, True]
    """
    # find DataFrame table...
    if (tbl_repr_type := _extract_table(data)) is not None:
        return _from_dataframe_repr(*tbl_repr_type)

    # ...or Series in the given string
    m = re.search(
        pattern=r"(?:shape: (\(\d+,\))\n.*?)?Series:\s+([^\n]+)\s+\[([^\n]+)](.*)",
        string=data,
        flags=re.DOTALL,
    )
    if m is not None:
        return _from_series_repr(m)

    msg = "input string does not contain DataFrame or Series"
    raise ValueError(msg)


def _from_dataframe_repr(tbl: str, table_repr: TableRepr) -> DataFrame:
    """Reconstruct a DataFrame from a table repr string."""
    from polars.datatypes.convert import dtype_short_repr_to_dtype
    from polars.io.database._inference import dtype_from_database_typename

    def _dtype_from_name(tp: str | None) -> PolarsDataType | None:
        return (
            None
            if tp is None
            else (
                dtype_short_repr_to_dtype(tp)
                or dtype_from_database_typename(tp, raise_unmatched=False)
            )
        )

    # associated regex patterns for the given table format

View on GitHub (pinned to df599052da)

Solutions

Make sure the string is the verbatim output of print(df) / repr(df) or repr(s) for a pl.Series, including the border rows and the 'shape: (n, m)' header
If the data crossed JSON/logs, repair escapes first (e.g. json.loads or codecs to unescape \\n) so newlines are real newlines
For durable round-trips, use real serialization instead: df.write_json / pl.read_json, write_ipc, or write_parquet
For a Series, confirm the string contains a 'Series:' line as produced by repr(series)

Example fix

# before
pl.from_repr("shape: (2, 2)\\n┌─────┐\\n│ a ─┐\\n...")  # double-escaped literal

# after
import json
pl.from_repr(json.loads('"shape: (2, 2)\\n┌─────┐\\n..."'))  # unescaped once, real newlines

Defensive patterns

Strategy: validation

Validate before calling

import re

def looks_like_polars_repr(s: str) -> bool:
    return bool(
        re.search(r'shape: \(\d+,? ?\d*\)\n', s)
        or re.search(r'Series:\s+[^\n]+\s+\[[^\n]+]', s)
    )

if not looks_like_polars_repr(text):
    raise ValueError('not a polars DataFrame/Series repr; refusing to call from_repr')

Try / catch

try:
    obj = pl.from_repr(text)
except ValueError as e:
    if 'does not contain DataFrame or Series' in str(e):
        # recover: try real serialization or surface a clear fixture error
        raise AssertionError(f'fixture string is not a polars repr: {text[:80]!r}') from e
    raise

Prevention

Never use from_repr as a persistence format; write_ipc/write_parquet/write_json instead
When embedding reprs in fixtures, paste verbatim console output including borders and 'shape:' header
Un-escape strings that traveled through JSON/logs (json.loads) before parsing
Pin the polars version on both producer and consumer so the table format matches

When it happens

Trigger: pl.from_repr('hello world') or any prose/random string; passing str(df) of a LazyFrame or Expression instead of a DataFrame; a repr string whose borders/separator rows were mangled by copy-paste, terminal soft-wrapping, or log truncation; a repr from a polars version with a different table format.

Common situations: Copy-pasting console output of a DataFrame out of logs/notebooks into a test fixture; piping repr text through tools that strip or wrap characters; using from_repr as a serialization format in round-trip tests where the string got escaped (e.g. double-escaped \n).

Related errors

AI-assisted analysis of pola-rs/polars@df599052da (2026-08-16). Data as JSON: /api/errors/5742f47d46fb964b. Report an issue: GitHub.