hide-vault-name

Hides every element that renders the vault’s NAME, so a captured frame does not depend on which temporary vault the run happened to get.

The harness names its temporary vaults temp-vault-<random>, and Obsidian prints that name in the vault-switcher row at the bottom of the left sidedock. A desktop capture therefore carries the random suffix, which makes the PNG differ on every run — measured as 349 differing pixels in x:153-197 y:774-783 of a 1200x800 frame, amplified 16x to read temp-vault- plus that run’s six-character suffix. The caption band labelScreenshot draws over that strip attenuates it to about 6 % brightness rather than hiding it, so it is invisible to a reader and fatal to reproducibility: a capture suite rewrites its checked-in screenshots on every run and the diff is unreadable.

captureObsidianScreenshot calls this by default, so a consumer gets a reproducible frame without knowing the helper exists. It is exported anyway, for a consumer capturing through a raw connectToCdp connection — the same reason measureLabelCaption is exported rather than kept private to the labeller.

Interfaces

Interface	Description
HideVaultNameOptions	Options for hideVaultName.

Functions

Function Description

hideVaultName Hides every element whose text is exactly the vault’s name. Keyed on app.vault.getName() rather than on a class of Obsidian’s chrome, deliberately: the varying thing is the DATA, and the row that renders it has moved between Obsidian versions before. Obsidian’s own hide() rather than an inline style.visibility or an injected stylesheet, because those are what obsidianmd/no-static-styles-assignment and obsidianmd/no-forbidden-elements respectively refuse. Only leaf elements match (childElementCount === 0), so an ancestor whose textContent happens to concatenate to the same string is left alone — on a vault named after the only note in it, hiding the ancestor would blank half the frame. Idempotent, and meant to be re-applied. hide() leaves the text in place, so a second call finds what the first one hid; and it has to, because a re-render of the sidedock between two captures puts the row back. Not asserted here. Zero is a legitimate answer — the mobile app renders no vault-switcher row at all — so the count is returned rather than checked. A desktop capture suite that wants to fail loudly when Obsidian relocates the row should assert on it itself.

Function	Description
hideVaultName	Hides every element whose text is exactly the vault’s name. Keyed on `app.vault.getName()` rather than on a class of Obsidian’s chrome, deliberately: the varying thing is the DATA, and the row that renders it has moved between Obsidian versions before. Obsidian’s own `hide()` rather than an inline `style.visibility` or an injected stylesheet, because those are what `obsidianmd/no-static-styles-assignment` and `obsidianmd/no-forbidden-elements` respectively refuse. Only leaf elements match (`childElementCount === 0`), so an ancestor whose `textContent` happens to concatenate to the same string is left alone — on a vault named after the only note in it, hiding the ancestor would blank half the frame. Idempotent, and meant to be re-applied. `hide()` leaves the text in place, so a second call finds what the first one hid; and it has to, because a re-render of the sidedock between two captures puts the row back. Not asserted here. Zero is a legitimate answer — the mobile app renders no vault-switcher row at all — so the count is returned rather than checked. A desktop capture suite that wants to fail loudly when Obsidian relocates the row should assert on it itself.