---
title: Formats
description: Text formats store their snapshots as UTF-8 without a byte order mark, with LF line endings and one trailing newline; images are stored exactly as given...
url: https://docs-v3-preview.elastic.dev/formats
---

# Formats
Text formats store their snapshots as UTF-8 without a byte order mark, with LF line endings and one trailing newline;
images are stored exactly as given. Each compares in its own way:

| Format                                                                  | Stored as                                                   | Compared                                             |
|-------------------------------------------------------------------------|-------------------------------------------------------------|------------------------------------------------------|
| [Text](https://docs-v3-preview.elastic.dev/formats/text)                | the text                                                    | line by line, shown as unified hunks                 |
| [Inline](https://docs-v3-preview.elastic.dev/formats/inline)            | a string literal in the test                                | as its format: text, HTML, JSON or YAML              |
| [HTML](https://docs-v3-preview.elastic.dev/formats/html)                | pretty-printed HTML                                         | as a DOM with AngleSharp.Diffing, plus a line diff   |
| [JSON](https://docs-v3-preview.elastic.dev/formats/json)                | pretty-printed JSON                                         | as a tree: keys in any order, numbers by value       |
| [YAML](https://docs-v3-preview.elastic.dev/formats/yaml)                | the YAML as written                                         | as a tree with the YAML 1.2 core schema              |
| [Images](https://docs-v3-preview.elastic.dev/formats/images)            | the PNG as given                                            | by pixels, perceptually, with anti-aliasing left out |
| [Browser journeys](https://docs-v3-preview.elastic.dev/formats/browser) | a transcript of a journey's reads                           | section by section, accessibility trees as trees     |
| [HTTP](https://docs-v3-preview.elastic.dev/formats/http)                | a transcript of each response: status, chosen headers, body | section by section, JSON bodies as trees             |

Any of the text formats can also live in the test itself, as an [inline snapshot](https://docs-v3-preview.elastic.dev/formats/inline).
Structural formats never rewrite a verified file that still matches: reformatting it by hand, a whitespace change in
the output, or an image re-encoded with the same pixels does not fail a test.

## Scrubbing volatile values

Values that change every run, such as GUIDs and timestamps, can be replaced before comparing:
```csharp
json.Should().MatchJsonSnapshot(options: new SnapshotOptions().Scrub(Scrubbers.Guids).Scrub(Scrubbers.DateTimes));
```

`Scrubbers.Guids` and `Scrubbers.DateTimes` number what they replace (`Guid_1`, `Guid_2`, …), the same value always
getting the same number, so references between values survive. A scrubber is any `Func<string, string>`, or a regex
and its replacement. To scrub every snapshot, register them once, for instance from a module initializer:
```csharp
[ModuleInitializer]
internal static void Snapshots() => SnapshotDefaults.AddScrubber(Scrubbers.Guids);
```

`Scrubbers.Path(root, replacement)` replaces a folder where it starts a path and writes the rest of the path with `/`, so a Windows run
gives the same text (`C:\work\repo\docs\a.md` and its JSON-escaped form both become `{repo}/docs/a.md`); `Scrubbers.TempDirectory()` is the
system temp folder as `{temp}`, including the resolved form where the temp folder is a link (`/private/var` on macOS):
```csharp
new SnapshotOptions().Scrub(Scrubbers.Path(repoRoot, "{repo}")).Scrub(Scrubbers.TempDirectory());
```

A drive letter from an in-memory file system (`C:\docs\docset.yml` against `/docs/docset.yml`) is not a root of this machine: scrub it with
`Scrubbers.Path("C:/docs", "/docs")`, or a regex.
Nothing is scrubbed unless you ask for it.