Skip to content

Logging & Timing

Monitor what additory is doing under the hood with operation logging and performance timing.


Operation Logging

Pass logging=True to any core function to see detailed operation logs:

import additory as add
import polars as pl

df = pl.DataFrame({
    'price': [100, 200, 300],
    'quantity': [2, 3, 1],
})

result = add.transform('@calc', df, expression='price * quantity',
    name='total', logging=True)

This prints log messages showing:

  • Function entry point and parameters
  • Validation steps
  • Rust operation calls
  • Data conversion steps

Logging uses Python's standard logging module at the INFO level.


Performance Timing

Set the ADDITORY_TIMING environment variable to get a detailed timing breakdown:

export ADDITORY_TIMING=true
python my_script.py
import os
os.environ['ADDITORY_TIMING'] = 'true'

import additory as add
# ... your code here

When enabled, each operation prints a timing breakdown:

======================================================================
ADDITORY TIMING BREAKDOWN (add.to)
======================================================================
  Arrow encode (ref):        0.12 ms
  Arrow encode (target):     0.08 ms
  Rust operation:            1.45 ms  ← Main operation
  Arrow decode:              0.09 ms
  ─────────────────────────────────
  Total:                     1.74 ms

  Rust operation: 83.3% of total time
  Arrow IPC:      16.7% of total time
======================================================================

This helps identify whether time is spent in:

  • Arrow IPC encoding/decoding — converting between Python DataFrames and Rust
  • Rust operation — the actual computation
  • Python overhead — validation, lineage tracking, type conversion

Supported Functions

Function logging ADDITORY_TIMING
add.to() ✅ ✅
add.transform() ✅ ✅
add.synthetic() ✅ ✅
add.scan() — —

Practical Usage

Debugging unexpected results

# See exactly what parameters are being passed to Rust
result = add.to(orders, customers, 'name', 'customer_id', logging=True)

Benchmarking

import os
os.environ['ADDITORY_TIMING'] = 'true'

# Compare performance of different approaches
result1 = add.transform('@calc', df, expression='x * 2', name='x2')
result2 = add.transform('@calc', df, strategy={'x2': 'x * 2', 'x3': 'x * 3'})

Disable timing in production

Set ADDITORY_TIMING=false (or unset it) to suppress timing output. The overhead of timing measurement itself is negligible.


Next Steps