Formula Language

Flowfile formulas are a simple, Excel-like expression language for transforming data. Write [column] to reference a column, call functions like round() or concat(), and use if ... then ... else ... endif for conditional logic — the same way you would in a spreadsheet formula or a SQL CASE statement.

Every formula compiles to a native Polars expression before it runs, so there is no row-by-row Python overhead: formulas are as fast as hand-written Polars code. The same language is used everywhere in Flowfile — the visual editor and the Python API share it.

Try it in your browser

The interactive formula playground. Pick a sample dataset, type a formula, and watch the result, the generated Polars code, and the FlowFrame code update as you type.

---

Where formulas are used

Surface	How
Formula node	Create or replace a column with a formula.
Filter node (advanced mode)	Keep rows where a formula evaluates to true; split mode routes passing and failing rows to separate outputs.
Python API	Pass formula strings to with_columns(flowfile_formulas=...), filter(flowfile_formula=...), and filter_split(flowfile_formula=...).

---

Syntax

Column references

Wrap column names in square brackets. Spaces in names are fine.

[salary]
[Order Date]
[first_name]

Literals

Strings use single or double quotes; numbers and booleans are written bare.

"hello"   'world'
42   3.14   -7
true   false

Operators

Operators	Meaning
+ - * / %	Arithmetic (% is modulo)
= == !=	Equality (= and == are equivalent)
> >= < <=	Comparison
and or	Boolean logic
( )	Grouping

Use parentheses to control evaluation order:

[price] * (1 - [discount])

Use and / or, not && / ||

Boolean logic uses the keywords and and or (case-insensitive). The C-style operators && and || are not part of the language.

Conditionals

if ... then ... else ... endif, with optional elseif branches:

if [score] >= 90 then "A"
elseif [score] >= 80 then "B"
else "C" endif

Functions

There are 95 built-in functions for logic, strings, math, dates, and type conversion — see the function reference. Calls can be nested.

uppercase(left([last_name], 3))
round([price] * [qty], 2)
coalesce([nickname], [name], "n/a")

Comments

Everything after // on a line is ignored.

[price] * [quantity]        // subtotal
  - ifnull([discount], 0)   // minus discount

---

Worked examples

Classify order volume and compute a discounted price:

if [quantity] > 75 then 'High' else 'Low' endif

round([price] * (1 - [discount]), 2)

Build a display name and compute tenure in days:

concat(titlecase([first_name]), " ", titlecase([last_name]))

date_diff_days(today(), [hire_date])

---

Function reference

Category	Functions	Examples
Logic & Nulls	13	coalesce, ifnull, between, is_empty
String	23	concat, uppercase, trim, replace, split
Math	23	round, abs, floor, power, log
Date & Time	28	year, add_days, date_diff_days, format_date
Type Conversion	8	to_string, to_integer, to_date, to_boolean

---

How it works

1. Parse — the formula is tokenized and each token is classified: columns, literals, operators, functions, and if/then keywords.
2. Build — the tokens are arranged into a function tree, applying operator precedence.
3. Convert — the tree is turned into a single Polars expression; execution is handled by Polars itself.

For example:

if [score] >= 90 then "A" elseif [score] >= 80 then "B" else "C" endif

becomes:

pl.when(pl.col("score") >= pl.lit(90)).then(pl.lit("A"))
  .when(pl.col("score") >= pl.lit(80)).then(pl.lit("B"))
  .otherwise(pl.lit("C"))

This is also why flows that use formulas export to clean Python code: each formula has a direct Polars equivalent.

---

Next: browse the function reference, use formulas in the Formula node, or call them from the Python API.