Code Editor

Code Editor (Advanced Mode) lets you write Python scripts directly inside your Transform pipeline. The dataframe is available as df. Polars, NumPy, SciPy, math, and re are preloaded - no imports required. Write your transformation below the read-only preamble, test it, then save.

Code Editor operations can be placed anywhere in a pipeline alongside no-code operations. The output of each step becomes the input for the next, regardless of whether the step is a code or no-code operation.

When to Use Code Editor

• Multi-condition logic. Classify rows using pl.when().then().otherwise() chains that no-code Filter cannot express.
• Custom column derivation. Create columns from arithmetic, string manipulation, or pattern matching across multiple existing columns.
• Regex extraction. Extract substrings, domains, or codes from text columns using re directly on string series.
• Numerical operations. Apply NumPy or SciPy functions to columns for statistical scoring, normalization, or scientific computation.
• Edge cases beyond no-code. Any transformation that requires logic branching, cross-column calculation, or library functions not available in the no-code operations.

Sample Dataset

The examples in this doc use the Edilitics sample orders dataset. Download it to follow along in your own workspace.

Editor Setup

When the Code Editor opens, the top of the editor contains a read-only preamble block. This block cannot be edited:

import re
import math
import scipy
import numpy as np
import polars as pl

# Transform your dataframe using Polars syntax.
# The dataframe object is available as: df
# Example: df = df.with_columns([df["column_name"].cast(pl.Int64)])

# Notes:
# - No imports are required or supported - Polars is preloaded.
# - Only whitelisted packages/functions are allowed.
# - Code runs on a 16,600-row preview sample only.

# Begin your script below:
#dataframe object = df

Write your transformation code below this block. The preamble handles all library loading. Reassign df to apply changes.

Restrictions

The editor blocks a set of Python keywords and patterns on every keystroke. Blocked terms trigger a "Not Allowed" toast and the change is rejected.

Write expression-level code using df = assignments. Do not define functions or use control flow keywords.

How to Use Code Editor

df = df.with_columns(
    pl.when(pl.col("revenue") > 1000)
      .then(pl.lit("High"))
      .when(pl.col("revenue") > 500)
      .then(pl.lit("Medium"))
      .otherwise(pl.lit("Low"))
      .alias("revenue_band")
)

Before and After

Rep performance score with percentile rank

Score each sales rep by combining their revenue share within region and discount efficiency - a multi-step calculation requiring window functions, cross-column arithmetic, and percentile ranking in a single pass. No combination of no-code operations can produce this in one step.

Input (first 3 rows of edilitics_sample_orders):

Script:

df = df.with_columns([
    (pl.col("revenue") / pl.col("revenue").sum().over("region")).alias("revenue_share_in_region"),
    (pl.col("revenue") / pl.col("units_sold")).alias("revenue_per_unit"),
])

df = df.with_columns(
    (
        pl.col("revenue_share_in_region") * 0.6
        + (1 - pl.col("discount_pct")) * 0.4
    ).alias("rep_score_raw")
)

df = df.with_columns(
    (pl.col("rep_score_raw").rank("average") / pl.col("rep_score_raw").len())
    .alias("rep_score_pct_rank")
)

Output (first 3 rows):

Output: 500 rows, 17 columns (13 original + 4 derived).

Example Scripts

Z-score normalisation per region

Normalise revenue within each region to identify outliers. Requires per-partition mean and standard deviation - not available in any no-code operation.

df = df.with_columns([
    pl.col("revenue").mean().over("region").alias("region_revenue_mean"),
    pl.col("revenue").std().over("region").alias("region_revenue_std"),
])

df = df.with_columns(
    ((pl.col("revenue") - pl.col("region_revenue_mean")) / pl.col("region_revenue_std"))
    .alias("revenue_zscore")
)

df = df.filter(pl.col("revenue_zscore").abs() < 3)

Extract structured fields from a JSON metadata column

The metadata column contains JSON strings like {"source": "web", "campaign": "q1_inbound"}. Extract both fields into separate columns. Requires row-level string parsing - not possible with String Extract or any no-code operation.

df = df.with_columns([
    pl.col("metadata").map_elements(
        lambda x: re.search(r'"source":\s*"([^"]+)"', x).group(1) if x and re.search(r'"source":\s*"([^"]+)"', x) else None,
        return_dtype=pl.String
    ).alias("lead_source"),
    pl.col("metadata").map_elements(
        lambda x: re.search(r'"campaign":\s*"([^"]+)"', x).group(1) if x and re.search(r'"campaign":\s*"([^"]+)"', x) else None,
        return_dtype=pl.String
    ).alias("campaign"),
])

Days to delivery and SLA breach flag

Compute the number of days between order_date and delivery_date, then flag rows where delivery took longer than 30 days. Requires date casting, cross-column arithmetic, and a derived boolean - not achievable in a single no-code step.

df = df.with_columns([
    pl.col("order_date").cast(pl.Date).alias("order_date_dt"),
    pl.col("delivery_date").cast(pl.Date).alias("delivery_date_dt"),
])

df = df.with_columns(
    (pl.col("delivery_date_dt") - pl.col("order_date_dt"))
    .dt.total_days()
    .alias("days_to_delivery")
)

df = df.with_columns(
    (pl.col("days_to_delivery") > 30).alias("sla_breach")
)

Cumulative revenue milestone flag per sales rep

Compute running total of revenue per sales rep ordered by date, then flag the row where cumulative revenue crosses 50,000 for the first time. Requires ordered cumulative sum partitioned by rep - not available in Window Functions no-code operation for milestone detection.

df = df.sort(["sales_rep", "order_date"])

df = df.with_columns(
    pl.col("revenue").cum_sum().over("sales_rep").alias("cumulative_revenue_by_rep")
)

df = df.with_columns(
    (
        (pl.col("cumulative_revenue_by_rep") >= 50000)
        & (pl.col("cumulative_revenue_by_rep").shift(1).over("sales_rep") < 50000)
    ).alias("crossed_50k_milestone")
)

Operation Reference

Frequently Asked Questions

Next Steps

Need help? Email support@edilitics.com with your workspace, job ID, and context. We reply within one business day.