Polars

Polars PanicException: Index Out of Bounds - How to Fix It

Answer

This Polars error means you're trying to access a row or element at a position that doesn't exist. Fix it by checking the DataFrame height with df.height before indexing, and use .head(), .tail(), or .slice() instead of direct row access when possible.

Why This Happens

Polars uses zero-based indexing. If your DataFrame has 5 rows (indices 0-4) and you try to access row 5, Polars panics. This commonly happens after filtering reduces row count, when using hardcoded indices, or when iterating beyond the actual data length.

Solution

The rule: always check df.height before indexing. Use .head(), .tail(), or .slice() for safer row access. After filtering, verify the result isn't empty before accessing rows.

import polars as pl

df = pl.DataFrame({
    'a': [1, 2, 3, 4, 5],
    'b': [10, 20, 30, 40, 50]
})

# ❌ Problematic: accessing row beyond bounds
df.row(10)
# PanicException: index out of bounds: 10 >= 5

# ✅ Debug: check DataFrame height first
print(f"Rows: {df.height}")  # 5
print(f"Valid indices: 0 to {df.height - 1}")  # 0 to 4

# ✅ Fixed: use valid index
df.row(4)  # last row

# ✅ Fixed: use negative indexing for last rows
df.row(-1)  # last row
df.row(-2)  # second to last

# ✅ Fixed: use slice for safe range access
df.slice(0, 3)  # first 3 rows
df.head(3)      # also first 3 rows
df.tail(2)      # last 2 rows

# ❌ Problematic: index after filtering
filtered = df.filter(pl.col('a') > 10)  # 0 rows!
filtered.row(0)
# PanicException: index out of bounds

# ✅ Fixed: check if DataFrame is empty
if filtered.height > 0:
    first_row = filtered.row(0)
else:
    print("No matching rows")

# ✅ Fixed: use get() pattern with default
def safe_get_row(df, idx):
    if 0 <= idx < df.height:
        return df.row(idx)
    return None

# ✅ For column access by index
df.select(pl.nth(0))  # first column
df.select(pl.nth(-1))  # last column

# ✅ Iterate safely
for i in range(df.height):
    row = df.row(i)
    # process row

Better Workflow

In Zerve, each block displays DataFrame height in its output. You see immediately that your filtered DataFrame has 0 rows before you try to access row(0). The visual flow shows where row counts drop, so you catch empty DataFrames at the filtering step, not when indexing fails downstream. No more off-by-one mysteries or hardcoded indices that worked yesterday but break today. Check the shape visually, then write your access logic with confidence.