Interaction design · Desktop utility
Radial menu inside an OS-wide drag
A Windows file converter with no window. Hold Shift while dragging any file and a radial menu of target formats opens under the cursor; release over a format and the file is converted in place, offline. It is built on a well-studied idea: in a pie menu, every option is the same distance from the pointer.
- Solo designer & developer
- Solo
- Independent · 2026
- Released & open source
01The problem
Converting a file is almost never the task; it is a step inside another task, like writing a report or preparing an upload. Yet the usual tools turn it into a detour: open an app, find the file, drop it on a target, pick a format from a list, choose where to save. Online converters shorten the detour but ask you to upload the file. The design question was how short the detour could get without leaving the place the file already is.
02Research & rationale
The idea comes from menu research rather than from users I studied. Fitts's law predicts that pointing time grows with distance and shrinks with target size, and in a linear menu the lower items are further away. Pie menus put every item at the same distance, and a controlled comparison found them faster, with fewer errors, than linear menus (Callahan et al., 1988). Marking menus then showed that once people learn an item's direction they can select it with a quick flick, without waiting for the menu to appear (Kurtenbach & Buxton, 1994). Mandarin puts that kind of menu inside an operating-system-wide drag. What I have not done yet is measure whether it works that way for anyone but me; that is the study below.
03Iterations
Concept A: a persistent floating drop panel
The first prototype was a small panel that stayed on the desktop and accepted dropped files. Using it myself, it still meant dragging a file across the screen, sometimes across monitors, to hit a small target. It removed the window, but not the travel.
Concept B: a Shift + drag radial wheel
So the menu comes to the file instead. A global hook watches for a drag with Shift held and opens the wheel centred on the pointer. Each format has a fixed direction (PNG to the upper left, WebP to the right), so with practice the choice can become a direction rather than a search. Releasing the mouse button commits it, with no extra click.
Screen 1 of 2
Screen 2 of 2
Progressive disclosure: Shift + Alt for advanced tools
I kept wanting crop, compression and metadata removal as well, but adding them to the main wheel would have crowded it and made its directions harder to learn; marking menus lose accuracy quickly as the number of items per level grows (Kurtenbach & Buxton, 1993). They live on a second wheel, opened with Shift + Alt, that uses the same gesture, so the primary wheel stays at eight formats.
Screen 1 of 4
Screen 2 of 4
Screen 3 of 4
Screen 4 of 4
04Screens
05Planned user study
Radial wheel vs. linear menu: a within-subjects timing study
Is choosing a target format on Mandarin's radial wheel faster and less error-prone than choosing it from a linear menu of the same formats, and does the difference hold on the first attempts, before any direction has become habit?
- Within-subjects, two conditions, with order counterbalanced across participants (half start with each). Same files, same conversion engine, same place on screen: only the shape of the menu changes.
- 8 to 12 university students who have never used Mandarin, recruited on campus. One session of about 25 minutes each.
- Radial: Shift + drag opens Mandarin's wheel of eight formats under the cursor
- Linear: Shift + drag opens a vertical list of the same eight formats, in the same order, at the same spot
- Five practice trials per condition, then 24 prompted conversions per condition (for example 'Convert chart.png to WebP'), each of the eight formats appearing three times in shuffled order. Only prepared dummy files are used, never anyone's own.
- Selection time, from the menu opening to the release, logged by Mandarin
- Errors: releases on the wrong format
- Learning: selection time across three blocks of eight trials
- System Usability Scale after each condition (Brooke, 1996)
- Raw NASA-TLX workload after each condition (Hart & Staveland, 1988)
- Preference, and two open questions at the end
- H1: By the final block, radial selection is faster than linear selection.
- H2: Radial selection produces no more errors than linear selection.
- H3: The radial advantage grows across blocks, as directions are learned.
- Per-participant medians compared with a Wilcoxon signed-rank test, because the sample is too small to assume normality, and reported with effect sizes rather than p-values alone. With 8 to 12 people this is a pilot: it can show a direction and an effect size worth following up, not settle the question.
- Written informed consent, anonymous participant codes, dummy files only, and every participant can stop at any time without giving a reason.
This study has not run yet, so there are no results here, only the plan. When it has run, the results will replace this note, whichever way they come out.
06Result
Mandarin is released and open source. How much time it actually saves is exactly what I have not measured yet. The study above is designed to answer that, including the case the research literature is least sure about for a tool like this: the first few uses, before a direction becomes habit.
- C#
- .NET 8
- WPF
- Win32 API
07Limitations
- All testing so far is my own. I know every direction on the wheel, which makes me the worst possible test user.
- Holding Shift during a drag is not discoverable on its own; the tool depends on people being told the gesture exists.
- The published speed advantage of pie menus comes from lab studies with a mouse; whether it survives inside a real drag, on a touchpad, is an open question.
08What I would do differently
I built the whole tool before timing a single selection. Next time I would prototype the wheel alone, with fake conversions, and time it against a plain list in the first week. The conversion engine was the expensive part, and it did not need to exist to test the interaction.
09References
- Fitts, P. M. (1954). The information capacity of the human motor system in controlling the amplitude of movement. Journal of Experimental Psychology, 47(6), 381 to 391. doi:10.1037/h0055392
- Callahan, J., Hopkins, D., Weiser, M., & Shneiderman, B. (1988). An empirical comparison of pie vs. linear menus. Proceedings of CHI '88. ACM. doi:10.1145/57167.57182
- Kurtenbach, G., & Buxton, W. (1993). The limits of expert performance using hierarchic marking menus. Proceedings of INTERCHI '93. ACM. doi:10.1145/169059.169426
- Kurtenbach, G., & Buxton, W. (1994). User learning and performance with marking menus. Proceedings of CHI '94. ACM. doi:10.1145/191666.191759
- Brooke, J. (1996). SUS: A 'quick and dirty' usability scale. In Usability Evaluation in Industry. Taylor & Francis.
- Hart, S. G., & Staveland, L. E. (1988). Development of NASA-TLX (Task Load Index). Advances in Psychology, 52, 139 to 183. doi:10.1016/S0166-4115(08)62386-9