Normalization Rules

All index keys are normalized before storage and lookup to ensure consistent matching.

Tags

• Lowercased: "#Project" → "#project"
• # prefix: ensured. Querying "project" or "#project" both work.
• Sources: getAllTags(cache) for combined, cache.tags for body, parseFrontMatterTags() for frontmatter.

Headings

• Lowercased: "My Heading" → "my heading"
• Heading level is not part of the key. # Foo and ## Foo produce the same key.

Frontmatter keys

• Lowercased: "Status" → "status"
• The position key from Obsidian’s internal FrontMatterCache is excluded.

Frontmatter values

YAML type	Normalization
String	Lowercased
Number	String(n) — e.g., 42 → "42"
Boolean	String(b) — true → "true"
Date (unquoted ISO)	toISOString().toLowerCase()
Array	Each element indexed separately
Nested object	JSON.stringify().toLowerCase()
null/undefined	Empty string ""

Composite key format: key\0normalizedValue (null byte separator).

See Frontmatter Value Types for how Obsidian stores different YAML types.

Aliases

• Lowercased: "Home Page" → "home page"

Backlinks

• Not lowercased: file paths are case-sensitive (matching TFile.path).
• Sources include non-markdown files (e.g., .canvas) from resolvedLinks.

Unresolved backlinks

• Lowercased: the raw link name is lowercased.

Embeds

• Subpaths stripped: "note#Heading" → resolved to "note.md" via getLinkpath() + getFirstLinkpathDest().
• See Embed Subpaths.

Block IDs

• No normalization: stored as-is. "myBlock" and "myblock" are distinct.

Task statuses

• No normalization: the raw character from ListItemCache.task is stored as-is. "x" and "X" are distinct.