Identifiers, display and revision
Jianlai tutorial 28: compare confusable glyphs without confusing appearance with characters
Use three fixed lines of Latin letters and digits to record legibility separately on keycaps, any available candidate area and host body text, then change just one real font-display setting. Deliver a character-identity card and layered observation table. Similar appearance does not mean identical code points, and no font is claimed inherently more accurate or medically beneficial.
On this page
Use three fixed lines of Latin letters and digits to record legibility separately on keycaps, any available candidate area and host body text, then change just one real font-display setting. Deliver a character-identity card and layered observation table. Similar appearance does not mean identical code points, and no font is claimed inherently more accurate or medically beneficial.
1. Establish identities independently of font shape
In the sample below, the first line is uppercase Latin I, lowercase Latin l and digit one. The second is uppercase O and digit zero. The third contains three adjacent letter-digit pairs. Write these identities on paper before examining the font so that every vertical-looking shape is not prematurely called one. Unicode’s Basic Latin chart lists distinct characters and names. Fonts can give them similar outlines without merging their encoded identities. Use this fixed nonprivate card, not real passwords, verification codes or accounts. Leave observation columns empty rather than predicting which font must be legible. Knowing code points is not evidence of a completed check on the current device.
2. Enter the fixed three-line case-sensitive sample
Confirm Jianlai in a private ordinary text draft and use actual case and digit controls to enter the literal sample. The first line has one ASCII space between each item; the second has one between its two items. The third has one between each pair, with no space inside a letter-digit pair. Commit composition and check case, especially that lowercase l has not been capitalized into a different character. Use two real newlines for three source lines and omit article fences. If the host auto-capitalizes a line start or substitutes text, record where it happens and restore the target in the copy. Do not replace a letter merely to imitate a shape, and assume no dedicated Jianlai code-point viewer.
I l 1
O 0
B8 S5 Z2
3. Map key code points to sample positions
On the external identity card, record uppercase I as U+0049, lowercase l as U+006C, digit one as U+0031, uppercase O as U+004F and digit zero as U+0030. Put these codes in the explanation, not in the three-line target. On line three, check uppercase B with eight, uppercase S with five and uppercase Z with two, preserving adjacency rather than interchanging characters that sound or look similar. If you already have a trusted offline character inspector, record its actual result on copied text separately. Without one, compare source, entered keys and body text item by item; the paper code-point chart is not an executed detection report. Distinct code points, similar outlines and ease of recognition are three separate judgments.
4. Separate keycaps, candidates and committed text
Hold theme, layout, host and screen state constant. First record which target letters and digits actually appear on keycaps, then whether the candidate area provides relevant content, and finally inspect the three committed body lines. If a character appears only on another layout, record that state rather than requiring every character on one screen. If no relevant candidate appears, mark the layer not applicable; do not invent candidates for every digit. Host body appearance also depends on the host’s own display conditions, so changing a keyboard font does not establish that document text changed. Record a specific confusing pair or line at each layer instead of a single overall quality score, enabling comparison at the same layer later.
5. Change one real display setting
Open an actual font or size setting in the installed version and make only one reversible change. Record the old value and theme, then use the real save or apply flow. If the current theme exposes no editable setting, record the limit and retain the baseline card rather than assuming arbitrary font-file import. Return to the same host with the same three source lines and inspect the three layers again. Do not also change background color, layout and host font. If only keycaps change while body text remains unchanged, record that by layer rather than calling it failure. An unchanged layer is not proof that saving failed; check the setting’s actual scope. The aim is controlled outline comparison, not a universally “best” font.
6. Distinguish a wrong character from an unclear glyph
Save a separate practice copy and replace only the middle lowercase l on the first line with digit one. Keep everything else unchanged. Identify the actual character change from the identity card, then compare the two appearances, which may differ strongly or only slightly. If they are hard to distinguish, the finding is poor distinguishability under this display condition, not identical characters. Restore lowercase l through the host selection or insertion point and check that neighboring uppercase I, digit one and both spaces remain unchanged. This counterexample gives separate evidence for appearance and source errors without real passwords or promises about candidates after an edit. Keep the correct reference untouched; the faulty copy is only for locating and restoring an error.
7. Diagnose case, entry and host font separately
If character identity is wrong, inspect the actual key, case state and host substitution first. If identity is correct but legibility poor, inspect the font and size at that particular layer. Keep those diagnoses separate. If pasted text looks different in another editor, check whether characters remain the same and whether the receiving host uses its own font. Without character evidence, retain uncertainty rather than claiming paste changed encoding. If the new display is absent after returning to Jianlai, verify the same extension and theme and restore the known old value if necessary. Change one condition per trial instead of cycling skins to hide an entry error. Preserve the three-line source while adding observations rather than overwriting earlier ones.
8. Deliver identities and layered observations
Deliver the correct three-line source, key code-point identity card, before-and-after observations by layer and the repaired-counterexample record. State which characters or candidate layers could not actually be compared and which settings had no available control; blank observations must not become universal passes. Restore the earlier readable configuration and reopen an ordinary input area, marking restoration only if checked. Any improved distinguishability is limited to this sample, device and host, not eyesight benefits, touch-hit accuracy or all text. Reuse the identity card for a future font, but observe again. Attractive shapes do not establish correct entered characters; identity and appearance evidence must remain separate.
Related guides and sources
Font and size comparison; Local revision and neighbour check.
Checked on 17 November 2026: Apple’s Finland lookup and China lookup both list Jianlai 1.0.5, requiring iOS / iPadOS 16.0 or later. The official store describes themes, backgrounds, fonts and multilingual input. Use controls actually present in the installed version; system editing and host processing are distinguished separately. These are repeatable instructions and paper expectations, not claims of completed device input, host compatibility or format-parser tests.
Method references: Unicode Basic Latin character chart; Apple text editing guide.