CEL Expression Language

The Computed Column transform evaluates a CEL expression against every row of your dataset. This page documents the expression syntax and the FileBender-specific extensions layered on top of it.

What is CEL?

CEL (Common Expression Language) is a non-Turing-complete expression language designed for speed and safety. FileBender uses the @marcbachmann/cel-js implementation, which covers most of the CEL specification.

Expressions are evaluated per-row. Each row is exposed as a flat object whose keys are the column names from your dataset.

Column references

There are two ways to reference a column value in an expression:

Dot syntax

When a column name is a valid identifier (letters, digits, underscores, starts with a letter), use dot notation directly:

revenue - cost

Bracket syntax

When a column name contains spaces, punctuation, or starts with a digit, wrap it in square brackets:

[Total Revenue] - [Operating Cost]

FileBender’s preprocessor rewrites bracket references into safe identifiers before the expression reaches the CEL engine. You can mix both styles in the same expression:

[Total Revenue] * tax_rate

Operators

Arithmetic

Operator	Description	Example
+	Addition	price + tax
-	Subtraction	total - discount
*	Multiplication	quantity * price
/	Division	total / count
%	Modulo	index % 2

FileBender registers cross-type arithmetic overloads so that numeric CSV columns (parsed as JavaScript floating-point numbers) work naturally with integer literals. An expression like Amount * -1 works even though Amount is a double and -1 is an int.

Comparison

Operator	Description	Example
==	Equal	status == "active"
!=	Not equal	status != "draft"
<	Less than	age < 18
<=	Less or equal	score <= 100
>	Greater than	amount > 0
>=	Greater or equal	count >= 10

Logical

Operator	Description	Example
&&	AND	age >= 18 && status == "active"
||	OR	role == "admin" || role == "owner"
!	NOT	!is_deleted

Other

Operator	Description	Example
? :	Ternary	score > 50 ? "pass" : "fail"
in	Membership	status in ["active", "pending"]

Built-in functions

Type conversion

Function	Description	Example
string(x)	Convert to string	string(123) → "123"
int(x)	Convert to integer	int("42") → 42
double(x)	Convert to floating-point	double("3.14") → 3.14
type(x)	Get the type name	type(42) → int
dyn(x)	Cast to dynamic type	dyn(42)

Collections

Function	Description	Example
size(x)	Length of list, string, or map	size("hello") → 5

String methods

These are called as methods on a string value:

Method	Description	Example
.contains(s)	Substring check	name.contains("Inc")
.startsWith(s)	Prefix check	code.startsWith("US")
.endsWith(s)	Suffix check	email.endsWith(".com")
.matches(regex)	Regex match	phone.matches("[0-9]+")
.size()	String length	name.size()
.indexOf(s)	First occurrence index	text.indexOf("@")
.lastIndexOf(s)	Last occurrence index	path.lastIndexOf("/")
.substring(start)	Slice from index	code.substring(2)
.substring(start, end)	Slice range	code.substring(0, 2)

Macros

Macro	Description	Example
has(field)	Check if a field exists	has(row.email)
list.all(x, cond)	All elements satisfy condition	scores.all(s, s > 0)
list.exists(x, cond)	Any element satisfies condition	tags.exists(t, t == "urgent")
list.exists_one(x, cond)	Exactly one satisfies	flags.exists_one(f, f == "primary")
list.map(x, expr)	Transform each element	values.map(v, v * 2)
list.filter(x, cond)	Keep matching elements	items.filter(i, i > 0)

Common patterns

String concatenation

first_name + " " + last_name

Conditional values

amount > 1000 ? "high" : "low"

Null-safe access

If a column might be missing or null, use has() to guard the access:

has(row.discount) ? price - row.discount : price

Numeric coercion

CSV values are parsed as JavaScript numbers (floating-point). Use int() if you need integer arithmetic:

int(quantity) * int(unit_price)

Combining string and number

name + " (qty: " + string(quantity) + ")"

Error handling

When a CEL expression fails for a given row (type mismatch, division by zero, missing variable), FileBender does not stop the flow. Instead:

• The failing row’s computed column value is set to null.
• The row is marked as an error in the execution results.
• Processing continues with the next row.

This means a single bad row won’t block the rest of your dataset. Check the execution results panel for error counts after running the flow.

Limitations

• Expressions are evaluated per-row. You cannot reference other rows or aggregate across the dataset within a single expression. Use Group By for aggregation.
• The CEL engine runs in the browser. Extremely complex expressions on large datasets may slow down processing.
• Only the CEL standard library subset shipped by @marcbachmann/cel-js is available. FileBender does not add custom functions beyond the arithmetic overloads described above.