> ## Documentation Index
> Fetch the complete documentation index at: https://docs.datafog.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# German entities

> Locale-gated German format and context detection, with exact source ranges.

<Note>
  This coverage is available in Core 0.4.0 or newer. All four runtimes
  share the same Core implementation.
</Note>

Activate them with `{"locale":"de"}`; `de-DE` and `de_DE` also work,
case-insensitively after trimming. Base detectors always continue to run.

## Scan and redact an IBAN

Pass the locale to scanning and the entity selection to transformations. This
example keeps the `IBAN:` label and replaces only the original account value.

<CodeGroup>
  ```rust Rust theme={null}
  use datafog_core::{
      scan_and_transform, ScanAndTransformConfig, ScanConfig,
      TransformationConfig, TransformationStrategy,
  };

  let transform = TransformationConfig::new(TransformationStrategy::Redact)
      .with_entities(vec!["DE_IBAN".to_owned()])
      .unwrap();
  let config = ScanAndTransformConfig::new(transform)
      .with_scan(ScanConfig::default().with_locale("de").unwrap());
  let result = scan_and_transform("IBAN: DE44 5001 0517 5407 3249 31", &config)
      .unwrap();

  assert_eq!(result.text, "IBAN: [DE_IBAN]");
  ```

  ```python Python theme={null}
  from datafog_core import scan_and_transform

  result = scan_and_transform("IBAN: DE44 5001 0517 5407 3249 31", {
      "scan": {"locale": "de"},
      "transform": {
          "default": {"strategy": "redact"},
          "entities": ["DE_IBAN"],
      },
  })

  assert result.text == "IBAN: [DE_IBAN]"
  ```

  ```javascript Node.js theme={null}
  import { scanAndTransform } from "@datafog/node";

  const result = scanAndTransform("IBAN: DE44 5001 0517 5407 3249 31", {
    scan: { locale: "de" },
    transform: {
      default: { strategy: "redact" },
      entities: ["DE_IBAN"],
    },
  });

  console.assert(result.text === "IBAN: [DE_IBAN]");
  ```

  ```javascript Browser/WASM theme={null}
  import { init, scanAndTransform } from "@datafog/wasm";

  await init();
  const result = scanAndTransform("IBAN: DE44 5001 0517 5407 3249 31", {
    scan: { locale: "de" },
    transform: {
      default: { strategy: "redact" },
      entities: ["DE_IBAN"],
    },
  });

  console.assert(result.text === "IBAN: [DE_IBAN]");
  ```
</CodeGroup>

For detection alone, pass `{"locale":"de"}` directly to `scan` (Rust:
`scan_with_config`). Omitted locale leaves German detection disabled. Recognized `en-US` and `fr` aliases retain base detection; unsupported explicit
locales raise a configuration error in Core 0.4.0. Selecting `DE_IBAN` for a transformation
does not activate detection, and transforming explicit findings requires no locale.

## Supported formats

| Label | Detector name | Value / context |
| - | - | - |
| `DE_IBAN` | `datafog-core/de-iban` | `DE` + 20 digits, compact or optional grouping `DE44 5001 0517 5407 3249 31`; no context |
| `DE_VAT_ID` | `datafog-core/de-vat-id` | `DE`, optionally one horizontal separator or hyphen, then nine digits; no context |
| `DE_TAX_ID` | `datafog-core/de-tax-id` | Eleven digits, optionally grouped 2/3/3/3; tax context below |
| `DE_SOCIAL_SECURITY_NUMBER` | `datafog-core/de-social-security-number` | Two digits, six digits, one ASCII letter, three digits; optional single horizontal separators between groups; insurance context below |
| `DE_POSTAL_CODE` | `datafog-core/de-postal-code` | `PLZ` + optional one horizontal separator, colon or hyphen + five digits; or `DE`/`D` + one horizontal separator or hyphen + five digits |
| `DE_PASSPORT_NUMBER` | `datafog-core/de-passport-number` | One ASCII letter and eight digits; passport context below |
| `DE_RESIDENCE_PERMIT_NUMBER` | `datafog-core/de-residence-permit-number` | `AT` and seven digits; residence context below |

All prefixes and labels match ASCII case-insensitively. Digits are `[0-9]`.
Horizontal separators are U+0020 SPACE, U+0009 TAB, U+00A0 NBSP and U+202F
NARROW NBSP. Optional value separators occur independently at group boundaries,
at most one each. IBAN separators after `DE`, arbitrary grouping, double
separators, hyphens, periods and newlines are rejected. A candidate immediately
preceded or followed by an ASCII letter/digit is rejected; adjacent non-ASCII
characters and punctuation are allowed. Longer identifiers cannot yield a
valid-length prefix.

## Required contexts

Context must immediately precede the value, separated by zero or more horizontal
separators, optionally one `:`, `#` or `-`, then zero or more horizontal
separators. The value still needs its ASCII-alphanumeric boundary. Thus
`IdNr.12345678901` works and `IdNr12345678901` does not. Context cannot start
inside an ASCII word or number, span a newline, or activate later values in a list.

* Tax: `SteuerID`, `Steuer-ID`, `Steuer ID`, `Steueridentifikationsnummer`,
  `Identifikationsnummer`, `IdNr`, `IdNr.`, `TaxID`, `Tax-ID`, `Tax ID`.
  The single spaces within the short labels may be any horizontal separator.
* Insurance: `Rentenversicherungsnummer`, `Sozialversicherungsnummer`, `RVNR`, `SVNR`.
* Passport: `Passnummer`, `Reisepass`, `Reisepassnummer`, `Passport`,
  `Passport No`, `Passport No.`, `Passport Number`. The spaces after `Passport`
  may be one or more horizontal separators.
* Residence: `Aufenthaltstitel`, `Aufenthaltserlaubnis`, `Residence Permit`, `eAT`.
  The space in `Residence Permit` may be one or more horizontal separators.

Tax, insurance, passport and residence findings include only the value, so
transformations retain their context. Postal findings include the prefix:
`PLZ:10115` is one complete finding; `PLZ: 10115` is deliberately rejected.
Structured keys and sibling values never supply context. Paths are JSON Pointers
and all ranges are local to the original string value.

## Detection policy and findings

There is no IBAN/VAT/tax checksum validation, bank or postal directory lookup,
pension date validation, document issuance validation, network call or model
requirement. Sensitive-looking transcription errors are intentionally detected.
Detection establishes neither validity nor existence of an account or identifier.
Passport and residence patterns preserve legacy heuristics and do not cover all
German documents. Personalausweis and tax-office Steuernummer are outside scope.

Every finding preserves original casing, internal separators and zero-based,
end-exclusive byte/code-point ranges. JavaScript bindings also expose UTF-16
ranges. Confidence is absent and detector version is the current crate version.
Repeated occurrences have separate source ranges. Scanning retains overlaps;
transformations resolve them using the existing selection rules and emit one
unnumbered placeholder such as `[DE_TAX_ID]` per selected finding.

## Structured values

The same locale applies independently within each string. Context must appear
in that string; a property named `Steuer-ID` does not supply textual context.

```python theme={null}
from datafog_core import scan_and_transform_structured

result = scan_and_transform_structured({
    "account": "DE44 5001 0517 5407 3249 31",
    "notes": ["Steuer-ID 12345678901"],
    "Steuer-ID": "12345678901",
}, {
    "scan": {"locale": "de"},
    "transform": {
        "default": {"strategy": "redact"},
        "entities": ["DE_IBAN", "DE_TAX_ID"],
    },
})

assert result.data == {
    "account": "[DE_IBAN]",
    "notes": ["Steuer-ID [DE_TAX_ID]"],
    "Steuer-ID": "12345678901",
}
```

Only German findings are selected in this example. Generic detectors may also
return overlapping findings; they remain available to callers. Configure
[overrides and allowlists](/guides/configuration) for the exact original values,
including their casing and separators.

See [German detection migration](/guides/migrating-from-datafog-python#german-detection-migration)
for intentional differences from Python 4.8.1 and the Python 4.9 release dependency.