Ligature
A technique that joins two or more characters and displays them as a single glyph. fi, fl, and similar pairs are typical examples.
A ligature is a typographic technique that joins two or more characters and displays them as a single glyph (character shape). In the Latin script, fi, fl, ff, ffi, and similar combinations are typical ligatures, and since the era of letterpress printing they have been used as a means of avoiding physical collisions between letters and achieving beautiful typesetting. This tradition carries on in modern digital fonts, and ligatures are widely implemented as a feature of OpenType fonts.
The mechanism of ligatures is defined in the GSUB (Glyph Substitution) table inside the font file. When the text rendering engine detects a specific sequence of characters, it draws the joined glyph instead of the individual glyphs. For example, when "f" and "i" are adjacent, the ascender of the f and the dot of the i interfere with each other, so replacing them with a dedicated ligature glyph keeps visual harmony.
There is more than one kind of feature that triggers the substitution. In OpenType feature tags, standard ligatures such as fi and ffi are assigned to liga, and standard ligatures applied depending on context to clig. Decorative ligatures prepared by the type designer are assigned to dlig, ligatures of old typesetting styles to hlig, and joins required by the writing system to rlig. In implementation, all of them are GSUB lookup type 4 (ligature substitution), which maps a sequence of multiple glyphs to 1 glyph. If you lump them together as "ligatures are the liga feature of a font," you will miss ligatures implemented with other features, such as those in the programming fonts described below.
In recent years, coding ligatures in programming fonts have been attracting attention. Fonts such as Fira Code, JetBrains Mono, and Cascadia Code display != as ≠, => as ⇒, and <= as ≤. This improves the visibility of operators and increases the readability of code. However, coding ligatures are only a display transformation, and the source code itself is saved as the original character sequence.
A point that is easy to be confused by in practice is that coding ligatures are implemented using calt (contextual alternates) rather than liga. Fira Code's official setup instructions guide you to enable them with font-feature-settings: "calt" or font-variant-ligatures: contextual. For this reason, specifying only no-common-ligatures does not remove the ligatures for arrows and comparison operators, and if you want to turn them off, you need to add no-contextual as well or specify none. If the editor has its own on / off setting for ligatures, that setting takes priority over CSS.
In CSS, you can control enabling / disabling ligatures with the font-variant-ligatures property. The values are divided into 4 groups: common-ligatures / no-common-ligatures correspond to liga and clig, discretionary-ligatures to dlig, historical-ligatures to hlig, and contextual to calt. What you should understand here is the behavior of the default value. With the initial value normal, common ligatures and contextual alternates are enabled, and discretionary ligatures and historical ligatures are disabled. In other words, if you specify a font that has ligatures, fi and ffi are applied without writing anything. You need to write CSS only "when you want to add decorative ligatures" or "when you want to stop ligatures." Specifying none disables all 4 groups.
A common misconception is confusing ligatures with kerning. Kerning is a technique that adjusts the distance between characters, whereas a ligature is a technique that replaces multiple characters with a single glyph. Both share the pursuit of beautiful typesetting, but their mechanisms are fundamentally different. Also, when joining letters is built into the basic rules of a writing system, as with the Arabic script, ligatures are an essential feature rather than decoration. A typical example is that lam followed by alef always takes the joined form, and OpenType registers this as rlig (Required Ligatures), treating the Syriac script in the same way. The specification states that control of this feature should generally not be exposed to users, so it is not something to turn off based on visual preference.
From the perspective of character counting, a ligature is drawn as a single glyph on screen, but internally it is treated as the original multiple characters. For example, an fi ligature looks like 1 character but is counted as 2 characters. When moving the cursor in a text editor, it is also common to be able to move 1 character at a time inside a ligature. The character count does not change when you change the font because this substitution happens only at the drawing stage.
However, there are exceptions. Unicode has code points that encode ligatures themselves as single characters, and for example fi (U+FB01 LATIN SMALL LIGATURE FI) and ffl (U+FB04) are counted as 1 character regardless of font features. They sometimes slip into text copied from old PDFs or word processor documents, and even though they look the same, they are different from the 2 characters f and i placed side by side. This difference surfaces in normalization. Passing them through compatibility normalization (NFKC) expands U+FB01 into the 2 characters f and i, and increases U+FB04 to 3 characters. On the other hand, composition normalization (NFC) does not change them, so in systems that assume NFC, they remain as 1 character. When the character count does not match your expectations, first check whether these code points are mixed into the input string and which normalization it passes through. Note that ß and Œ originated as ligatures but are treated as independent characters and are not decomposed even by NFKC.