Another faulty assumption is the idea that characters are laid out in a neat linear progression running in lines from left to right. In many languages, this isn't true.
Many languages also employ diacritical marks that are used in combination with other characters to indicate pronunciation. Exactly where the marks are drawn can depend on what they're being attached to. For example, look at these two letters:
Each of these examples is the letter a with an umlaut placed on top of it. The umlaut needs to be positioned higher when attached to the capital A than when attached to the small a.
This positioning can be even more complicated when multiple marks are attached to the same character. In Thai, for example, a consonant with a tone mark might look like this:
If the consonant also has a vowel mark attached to it, the tone mark has to move out of the way. It actually moves up and becomes smaller when there's a vowel mark:
Mark positioning can get quite complicated. In Arabic, a whole host of dots and marks can appear along with the actual letters. Some dots are used to differentiate the consonants from one another when they're written cursively, some diacritical marks modify the pronunciation of the consonants, vowel marks may be present (Arabic generally doesn't use letters for vowelsthey're either left out or shown as marks attached to consonants), and reading or chanting marks may be attached to the letters. In fact, some Arabic calligraphy includes marks that are purely decorative. There's a hierarchy of how these various marks are placed relative to the letters that can get quite complicated when all the various marks are actually being used.
Unicode expects that a text rendering process will know how to position marks appropriately. It generally doesn't encode mark position at allit adopts a single convention that marks follow in memory the characters they attach to, but that's it.3
Diacritical marks are not the only characters that may have complicated positioning; sometimes the letters themselves do. For example, many Middle Eastern languages are written from right to left rather than from left to right (in Unicode, the languages that use the Arabic, Hebrew, Syriac, and Thaana alphabets are written from right to left). Unicode stores these characters in the order they'd be spoken or typed by a native speaker of one of the relevant languages, known as logical order.
Logical order means that the "first" character in character storage is the character that a native user of that character would consider "first." For a left-to-right writing system, the "first" character is drawn farthest to the left. (For example, the first character in this paragraph is the letter L, which is the character farthest to the left on the first line.) For a right-to-left writing system, the "first" character would be drawn farthest to the right. For a vertically oriented writing system, such as that used to write Chinese, the "first" character is drawn closest to the top of the page.
Logical order contrasts with visual order, which assumes that all characters are drawn progressing in the same direction (usually left to right). When text is stored in visual order, text that runs counter to the direction assumed (usually the right-to-left text) is stored in memory in the reverse of the order in which it was typed.
Unicode doesn't assume any bias in layout direction. The characters in a Hebrew document are stored in the order they are typed, and Unicode expects that the text rendering process will know that because they're Hebrew letters, the first one in memory should be positioned the farthest to the right, with the succeeding characters progressing leftward from there.
This process gets really interesting when left-to-right text and right-to-left text are mixed in the same document. Suppose you have an English sentence with a Hebrew phrase embedded into the middle of it:
Even though the dominant writing direction of the text is from left to right, the first letter in the Hebrew phrase () still goes to the right of the other Hebrew lettersthe Hebrew phrase still reads from right to left. The same thing can happen even when you're not mixing languages: In Arabic and Hebrew, for example, even though the dominant writing direction is from right to left, numbers are still written from left to right.
This issue can be even more fun when you throw in punctuation. Letters have inherent directionality; punctuation doesn't. Instead, punctuation marks take on the directionality of the surrounding text. In fact, some punctuation marks (such as the parentheses) actually change shape based on the directionality of the surrounding text (called mirroring, because the two shapes are usually mirror images of each other). Mirroring is another example of how Unicode encodes meaning rather than appearancethe code point encodes the meaning (“starting parenthesis”) rather than the shape (either “(” or “)” depending on the surrounding text).
Dealing with mixed-directionality text can become quite complicated, not to mention ambiguous, so Unicode includes a set of rules that govern just how text of mixed directionality is to be arranged on a line. The rules are rather involved, but are required for Unicode implementations that claim to support Hebrew or Arabic.4
The writing systems for the various languages used on the Indian subcontinent and in Southeast Asia have even more complicated positioning requirements. For example, the Devanagari alphabet used to write Hindi and Sanskrit treats vowels as marks that are attached to consonants (which are treated as "letters"). A vowel may attach not just to the top or bottom of the consonant, but also to the left or right side. Text generally runs left to right, but when a vowel attaches to the left-hand side of its consonant, you get the effect of a character appearing "before" (i.e., to the left of) the character it logically comes "after."
In some alphabets, a vowel can actually attach to the left-hand side of a group of consonants, meaning this "reordering" may actually involve more than just two characters switching places. Also, in some alphabets, such as the Tamil example we looked at earlier, a vowel might actually appear on both the left- and right-hand sides of the consonant to which it attaches (called a "split vowel").
Again, Unicode stores the characters in the order they're spoken or typed; it expects the display engine to do this reordering. For more on the complexities of dealing with the Indian scripts and their cousins, see Chapter 9.
Chinese, Japanese, and Korean can be written either horizontally or vertically. Again, Unicode stores them in logical order, and the character codes encode the semantics. Many of the punctuation marks used with Chinese characters have a different appearance when used with horizontal text than when used with vertical text (some are positioned differently, some are rotated 90 degrees). Horizontal scripts are sometimes rotated 90 degrees when mixed into vertical text and sometimes not, but this distinction is made by the rendering process and not by Unicode.
Japanese and Chinese text may also include annotations (called "ruby" or "furigana" in Japanese) that appear between the lines of normal text. Unicode includes ways of marking text as ruby and leaves it up to the rendering process to determine how to draw it. For more on vertical text and ruby, see Chapter 10.