# ColumnMapping
## Properties
Name | Type | Description | Notes
------------ | ------------- | ------------- | -------------
**source_column** | **String** | Where the value comes from in the source record. A plain identifier (`user_id`) for a relational source, or a dotted path (`profile.address.city`) reaching inside a structured value for a document source. A single-segment path is exactly the bare column name this field has always carried, so pre-existing mappings are unchanged. Every segment must be a valid identifier; JSON Pointer syntax and array indices are rejected. A connector that only reads top-level fields (Postgres) refuses a dotted path as a dry-run precondition failure rather than at run time -- see `Capabilities.nested_source_paths`. |
**target_column** | **String** | |
**transform** | Option<**String**> | Optional restricted-CEL expression producing the value written to `target_column`. Omit it for a verbatim copy, which is what every mapping did before this field existed. Inside the expression `value` is this entry's own source value and `row.<column>` reads another column of the same row -- `row` may only be indexed by a literal column name, so the set of columns an expression reads is known before the run starts, fetched, and checked against the source schema. Deliberately restricted: an explicit allow-list of operators and functions (`size`, `int`, `double`, `string`, `bool`, `has`, `startsWith`, `endsWith`, `contains`, `lowerAscii`, `upperAscii`, `trim`, `replace`, `split`, `join`, `substring`), no I/O, no clock, no regular expressions, no iteration macros, no implicit type coercion, and both an evaluation-step and an output-size budget. Anything outside the allow-list is rejected at validation time and appears as a `MappingIssue` in the dry-run report, before any write. | [optional]
[[Back to Model list]](../README.md#documentation-for-models) [[Back to API list]](../README.md#documentation-for-api-endpoints) [[Back to README]](../README.md)