hl-valdemar/tubular
1
0
Fork 0
Code Issues Pull Requests Actions Packages Releases Wiki Activity

fix: match title animation groups by terminal width

Browse Source
This commit is contained in:
Valdemar Høgsvig Lorenzen
2026-10-19 18:55:38 +02:00
parent a44d95bb71
commit 5c4c387e5c
2 changed files with 185 additions and 38 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+1 -1
View File
@@ -46,7 +46,7 @@ After selection, the browser starts in **normal mode**. Press `i` to type a sear
Both providers use the same layout: a cyan view heading, bordered search and results, a bordered status line, and a dim one-line key hint bar. Entries use two lines: the title above dimmed metadata. YouTube shows the channel and publication date beneath each video title, without separate channel/date columns; tagged channels remain cyan. The selected row uses black text on cyan with a `> ` marker; saved titles are green, watch-later titles magenta, and `✓` marks watched entries. Metadata and help open in centered cyan-bordered popups; bulk-removal confirmations use yellow borders. Miruro keeps its catalog navigation and title/episode metadata.
YouTube's dimmed metadata row includes duration as `M:SS` or `H:MM:SS`, or `Unknown duration` when unavailable. Duration is retained in saved collections; older entries without duration remain readable.
YouTube keeps fast flat listings, then resolves visible videos' original-preferred titles asynchronously, one metadata request at a time. Titles replace the listing text left to right, one user-perceived character about every 20 ms; shorter titles erase the old suffix with spaces before dropping that padding. Offscreen changes apply immediately, and selection stays on the same video. Results are cached by video ID for the session; failures keep the listing title without automatic retries. This adds background yt-dlp requests, and opening video info can make a separate metadata request. “Original-preferred” uses the full extractor response, not a guarantee that YouTube will never return localized text.
YouTube keeps fast flat listings, then resolves visible videos' original-preferred titles asynchronously, one metadata request at a time. Titles replace the listing text left to right at 20 ms per terminal column, using the smallest groups of whole user-perceived characters with matching column widths. Two narrow characters and one wide character therefore swap together after 40 ms, without shifting the untouched suffix or inserting temporary blanks. Only an exhausted shorter title is padded with trailing spaces; that padding disappears when the animation finishes. Offscreen changes apply immediately, and selection stays on the same video. Results are cached by video ID for the session; failures keep the listing title without automatic retries. This adds background yt-dlp requests, and opening video info can make a separate metadata request. “Original-preferred” uses the full extractor response, not a guarantee that YouTube will never return localized text.
Selecting a YouTube video launches: