The Output Formatter

TADS 3 is in many ways a general-purpose programming system, but it is nonetheless designed above all else as a system for text-based interactive fiction. It might be tautological to say this, but one of the most important elements of text IF is the text. Any system for creating text-based IF had therefore better provide some good text formatting tools.


The display language of TADS 3 is HTML. Nearly, anyway: the HTML used in TADS is based on the official web standard for HTML 3.2, but a few features of HTML 3.2 are missing, a few things act differently than they do in regular HTML, and a number of extensions are available. The missing and modified features are mostly in areas where HTML features that were designed for web browsing just don't fit into text IF, or don't fit as they were designed in HTML. Where TADS adds new features to HTML, we've tried to follow later versions of the HTML standards wherever possible, or at least to stay consistent with the style of similar HTML features. Full details on the special version of HTML that TADS 3 uses can be found in the HTML TADS documentation, which describes all of the deviations from standard HTML.

HTML is used in all text output windows, including the main game window and all text banners. HTML is not used in "text-grid" banners, though.

Word Wrapping/Line Breaking

All of the interpreters perform automatic "word wrapping" on the output text, which means that when the text output would overflow the right margin, the interpreter finds the nearest word boundary and breaks the line there. This allows the author to largely ignore the details of how wide the game window is and how the text is laid out on the screen; the game simply writes out its paragraphs of text, and the system automatically makes it look right.

In some cases, though, it's desirable for the game to be able to control exactly how the system decides where to break lines. TADS 3 provides a sophisticated set of controls that give the game a great deal of detailed control over line breaking. Because of the substantial differences among the interpreters, it would be too difficult to force the game to actually do the line breaking itself-there would be too many cases to deal with. Instead, TADS 3 lets the game control line breaking with special advisory sequences that can be placed into the text itself. The interpreter determines where to break lines by inspecting the text, taking into account the advisory sequences.

(Note that the GUI text-only interpreters do not currently support the different wrapping modes, and do not support the override controls. These interpreters support only the traditional word-wrapping mode, so games that depend on the custom line-breaking features might not produce acceptable results on these systems.)

Word wrap and character wrap modes

To start with, TADS 3's output formatter has two basic modes: word wrapping and character wrapping.

In word-wrapping mode, the interpreter will only break a line between words. This is the default mode, and is the style that almost always applies to languages like English that use groups of letters to represent words. The formatter's rules for determining word boundaries are simple: a word boundary is a space, or a hyphen (but a break can occur only to the right of a hyphen, and only when the hyphen isn't followed by another hyphen).

In character-wrapping mode, the interpreter can break a line anywhere. This mode is especially applicable to languages like Chinese in which each character represents an entire word. In Chinese, convention allows line breaks to occur almost anywhere; each glyph is a separate word, so there is no need to keep most pairs of adjacent glyphs together on one line.

The wrapping mode is selected using the <WRAP> tag (which is a TADS extension, not a standard HTML tag). <WRAP WORD> sets word-wrapping mode, and <WRAP CHAR> sets character-wrapping mode. The interpreter always starts in word-wrapping mode. This tag is not a container, but simply an in-line mode switch, so there is no </WRAP> tag; to change out of the current mode, simply use another <WRAP> tag with the new mode.

Overriding the Default Breaking Rules

The default line-breaking rules - in both word-wrap and character-wrap modes - are very simple, but not adequate for every situation. For example, a game written in English might have some words that include hyphens that should always be kept together on one line, or a word that includes some other embedded punctuation that could serve as line-break points if needed. As another example, a game written in Chinese wouldn't really want line breaks to occur just anywhere; the conventions for line-breaking in Chinese actually require that certain sequences of characters, such as certain groups of punctuation marks, be kept together, and don't allow line breaks to occur just before or after certain types of punctuation.

It would be impossible to anticipate every possible line-breaking rule for every language and every game, so the interpreter doesn't even try; instead, the interpreter uses the rather simple set of rules outlined above by default, but provides a set of special control sequences that allow the game to override the default behavior whenever it wants. These special controls are described below.

The Zero-Width Non-Breaking Space

The control that prevents a line break is called the "zero-width non-breaking space." This is a special character, defined in the Unicode standard, that can be written in TADS as "\uFEFF" (that "\u" is the TADS way of entering a specific Unicode character code as a hexadecimal number); it can also be written with an HTML markup, &zwnbsp;. It's a "zero-width" character, which means that it doesn't show up on the display: it's simply invisible as far as the user is concerned. The "non-breaking" part is the special feature: this tells the interpreter that it cannot break the line here, even if it otherwise could.

To be more specific, the zero-width non-breaking space prevents a line break from occurring between the two characters adjacent to it. Essentially, this control is a bit of glue that sticks so strongly to the characters on both sides that the line breaking rules can't tear them apart.

Note that the glue only sticks to one side of each adjacent character. So, if you put a \uFEFF character just before a space, and you're in word-wrapping mode, the formatter can still break the line after the space. If you want to prevent the space from being a line break at all, you have to put a \uFEFF character on both sides of the space-one before and one after.

The Zero-Width Space

The other main control is the "zero-width space," and it does essentially the opposite of the zero-width non-breaking space: this character enables a line break where the rules would otherwise not allow it. This is another standard Unicode character, written in TADS as "\u200B" or with the HTML markup &zwsp;. Like \uFEFF, this is a zero-width character, so it's invisible on the display. Otherwise, though, it counts as a space-which means that the formatter, even in word-wrap mode, is free to break the line between the two adjacent characters, as though they were separated by an ordinary space.

You can use a zero-width space to add your own rules about where lines can be broken. For example, suppose your game has a bunch of words where you're using an equals sign as though it were a hyphen, and you want the formatter to be able to break to the right of these equals signs just like it would for hyphens. To do this, you would simply insert a \u200B character immediately after each of these equals signs; this would tell the interpreter that it can break the line just after the hyphen if necessary.

The Non-Breaking Space

In addition to the zero-width non-breaking space, Unicode defines a regular-width non-breaking space. This is written as "\u00A0", or with the HTML markup &nbsp;. This type of space looks exactly the same as an ordinary space (the kind you get by pressing the space bar on the keyboard), but it behaves as though it were a non-space character: the formatter never breaks a line at a non-breaking space (thus the name), it never combines a non-breaking space with adjacent ordinary spaces, and it never trims non-breaking spaces from the beginning or end of a line. For line-breaking purposes, the non-breaking space behaves as though it were an alphabetic character.

Apart from the obvious difference in visual size, the non-breaking space differs from the zero-width non-breaking space in that the zero-width version is like glue-it prevents a line break from being inserted between the adjacent characters. The ordinary non-breaking space, however, merely acts like a non-space character; if the line-breaking rules allow it, the formatter can break the line immediately before or after an ordinary non-breaking space.

The Soft Hyphen

Another special control, the "soft hyphen," lets the game tell the interpreter that it can break a word with hyphenation at a particular point, but that it doesn't have to. The soft hyphen is another standard Unicode character, "\u00AD", or equivalently in HTML, &shy;.

Soft hyphens are normally invisible, so you can freely insert them into words without adding visual clutter. When the formatter decides to take advantage of a soft hyphen to break a line, though, the soft hyphen is displayed as a normal hyphen at the end of the line.

Whitespace Combining

The output formatter automatically combines runs of space characters. Any time a series of consecutive space characters appear together, the formatter combines them all into a single space.

This feature is useful because it frees the game of concerns about too much or too little space when generating text. IF games often generate text in pieces, stringing together a series of pre-written sentences and phrases. Sometimes, it's difficult to know in advance how a particular fragment will be combined with other fragments, which makes it hard to anticipate exactly how many spaces there will end up being when everything is put together. The formatter's automatic combination of consecutive spaces simplifies this process greatly by allowing the author to throw in excess spaces wherever there's any uncertainty about the need for extra spaces; since the formatter will automatically get rid of any extra spaces that aren't actually needed, the text as finally displayed will look right in any case.

The formatter does not combine any of the typographical spaces with one another, because the game is asking for a specific amount of spacing when it uses these characters. However, when an ordinary space (the kind you get by pressing the space bar on the keyboard)-or a run of ordinary spaces-is adjacent to a typographical space, the formatter removes the ordinary space or spaces and keeps only the typographical space. For example, if the game writes the string " \u2002 \u2003 " (three ordinary spaces, an en space, two more ordinary spaces, an em space, and three more ordinary spaces), TADS will display only an en space followed by an em space: all of the runs of ordinary space are removed, since they're all adjacent to typographical spaces.

Newline Combining

Just as the formatter combines each run of ordinary spaces into a single space, it combines each run of newline ("\n") sequences into one line break. This feature is similar in purpose to whitespace combining, in that it allows the game to display "\n" sequences wherever it wants a line break, without worrying about whether some other piece of code just displayed another line break or is about to display another one.

Because "\n" sequences are combined, it's not possible to use "\n" to display an entirely blank line. To display a blank line, use the "\b" sequence: this ends any current line, just like "\n", but then displays a blank line. The formatter does not combine "\b" sequences: if you want two blank lines, just write out "\b\b".

Whitespace Trimming

Just as the output formatter consolidates each run of ordinary spaces down to a single ordinary space, the formatter completely eliminates (or "trims") spaces at the beginning and end of a line of text.

At the start of a line, the formatter trims only ordinary spaces. Typographical spaces at the start of a line are preserved exactly as given.

Note that during line breaking, the formatter always breaks at the rightmost edge of a run of spaces (ordinary or typographical). This means that line breaking always results in trailing spaces (spaces at the end of a line), never leading spaces (spaces at the start of the next line). The only way to put spaces at the start of a line is to put them after an explicit line break (such as a "\n" or "\b" character, or a

At the end of a line, the formatter trims all trailing spaces, including typographical spaces. However, the formatter does not trim any trailing spaces that precede any sort of non-breaking space (zero-width or ordinary). If you want trailing spaces to appear, simply put a zero-width non-breaking space after the last space you want to preserve as a trailing space. (In most cases, it simply doesn't matter whether or not trailing spaces are trimmed, because spaces at the end of a line are usually invisible-they're just extra blank pixels, after all. However, trailing spaces can be visible under certain conditions, such as when the text is underlined or has a background color different from the window's prevailing background color.)

TAB Alignment

The character-mode text-only interpreters and the multimedia interpreters provide support for the HTML 3.0 <TAB> tag. This can be used for fairly complex alignment effects, including centering text or aligning it at the right edge of the window, and simple data tables. Refer to the HTML TADS documentation for full details on how to use <TAB>.

Since the <TAB> tag is usable on both character-mode and multimedia interpreters, it's usually preferable to use <TAB> instead of <TABLE> for simple alignment. <TABLE> is considerably more powerful, since it provides both horizontal and vertical alignment capabilities, and does a great deal of complex layout work automatically; but <TABLE> is only available on the multimedia interpreters, so using it can hurt a game's usability on other interpreters. If you must use <TABLE> (and there often is no substitute), you should consider providing an alternative format for character-mode platforms.

The GUI text-only interpreters do not support either <TAB> or <TABLE>; it's not possible to create any sort of text alignment effects on these interpreters. This is due to the use of proportional type without full HTML support; eventually, these interpreters should be superseded by full HTML versions, so this should be a short-term limitation.

Typographical Spaces

In most computer text applications, there's just one kind of "space" character: the kind you get when you press the space bar on the keyboard. This ordinary space character is the one-size-fits-all visual separator to fill in the spaces between words, sentences, and everything else.

In traditional typography, printers (the people, not the computer peripherals) use a whole range of spacing sizes, with spaces of particular sizes for particular situations. The differences in space sizes tend to be pretty subtle, so the one-size-fits-all world of simple computer typography is a good enough approximation for most people and for most informal material. However, the finer gradations of spaces used in traditional typography are often indispensable for fine-tuning a layout or for creating special effects that aren't possible with ordinary space-bar spaces.

In addition to the ordinary space character (the kind you get when you press the space bar on your computer's keyboard), the Unicode standard defines a number of special space characters for these sorts of typographical effects, and TADS 3 supports many of these.

Naturally, character-mode platforms can't display spaces at arbitrary sizes; they're stuck with just one kind of space. TADS 3 accommodates character-mode by approximating the typographical spaces as a number of ordinary spaces. The wide spaces turn into two or three ordinary spaces, the thinner spaces turn into one space, and the very thin spaces turn into zero spaces. The approximation is meant to provide results that look natural for character-mode renditions; fortunately, this sort of mapping is long established thanks to typewriters and early computer printers, so TADS 3 simply tries to use the same conventions that people have long been using for typewritten text.

The GUI text-only interpreters could in principle provide true proportional renditions of the typographical spaces, but at the moment they don't; they simply use the same approximations as the character-mode interpreters. (If these interpreters come back into widespread use at some point, then someone might become motivated to enhance them proper support for typographical spaces; at the moment, this doesn't seem likely.)

The table below shows the supported typographical space characters and how they're displayed on the different interpreters.

Name Unicode HTML Multimedia Appearance Text-Only/Fixed-Pitch Appearance Comments
en space\u2002&ensp; Half an em space Two spaces This amount of space is sometimes used after sentence-ending punctuation (such as a period or question mark).
Em space\u2003&emsp; Depends on the font; usually equal to the font's point size in width Three spaces
Three-per-em space\u2004&tpmsp; One-third the width of an em space Two spaces
Four-per-em space\u2005&fpmsp; A quarter of an em space One space In most fonts, the ordinary space is about this size.
Six-per-em space\u2006&spmsp; One-sixth of an em space One space
Figure space\u2007&figsp; Equal in width to a digit zero ("0") One space
Punctuation space\u2008&puncsp; Equal in width to a period (".") None (zero spaces)
Thin space\u2009&thinsp; One-fifth of an em space One space
Hair space\u200A&hairsp; One-eighth of an em space None (zero spaces) This can be used to separate punctuation marks that would otherwise run together, such as a double quote from a nested single quote.
Quoted space"\ " Same as ordinary space Same as ordinary space

Note that the "text-only" appearance described in the table applies to both text-only interpreters and fixed-pitch fonts in multimedia interpreters. The multimedia interpreters use the text-only version of the spacing for fixed-pitch fonts to preserve the fixed-pitch nature of the fonts; using true proportional spacing for fixed-pitch type would throw off the uniform character alignment of these fonts.

Interpreter Classes

There are three different main types, or "classes," of interpreter. You can use any interpreter to run any game, so there's no need to compile your game specially for different interpreters; the different types of interpreters do have some differences in the way they display output, though.

The interpreter classes are:

Despite the differences, most games can be written without any special code to handle the different interpreter types. This is possible because all of the interpreters understand HTML; the text-only interpreters, which only handle a subset of HTML, silently ignore any HTML markups that they can't use. Ignoring HTML markups is often not as bad as it might at first appear. In many text-based IF games, the more advanced HTML features, while very useful to create the particular appearance the author wants, are to some extent window dressing; so when the game is run on a text-only interpreter, some of the visual appeal of the game might be lost, but the essence is preserved and the game remains fully playable. Some people even prefer the purer, most subdued appearance of the text-only display style.

In some cases, though, a game uses HTML for more than just fancy text effects, and can't afford to have the HTML ignored. In these cases, the game can check at run-time to determine the capabilities of the interpreter, and use an alternative display style when running on a text-only interpreter. This approach is a little more work than coding a single set of HTML that produces acceptable results on all interpreters, but it allows the game to take full advantage of the HTML features when available, while still providing a working game for text-only platforms. For example, a game might want to use a set of hyperlinks to implement a menu when running on an multimedia interpreter, so that players can click on links to select actions; but because the text-only systems don't support hyperlinks, the game could check the interpreter class and provide a numbered list of options when running on a text-only system. Or, a game might want to use pictures or sound effects to convey some key information in the story; when running on a text-only system, where these can't be displayed, the game could instead provide the information with some added text that doesn't appear when the game is played with a multimedia interpreter.

To determine the interpreter class, a game can call systemInfo(SysInfoInterpClass). This returns SysInfoIClassText for a character-mode text-only system, SysInfoIClassTextGUI for a GUI text-only system, and SysInfoIClassHTML for a full multimedia interpreter. (See the tads-io function set for details.)

Caps/no-caps flags

There are two special string sequences that let you control capitalization of the next character displayed. These are handy in situations where you want to write "boilerplate" text that plugs in substitution parameters. If you're writing a message that plugs in a value at the start of a sentence, for example, you'll want to make sure the first character of the substituted text is capitalized.

The \^ sequence is the "caps flag." When you display this in a text window, it causes the next character displayed to be converted to upper-case.

The \v sequence is the "no-caps flag." When you display this in a text window, it causes the next character displayed to be converted to lower-case.

The output formatter tries to handle the "caps" and "no-caps" flags smoothly when combined with HTML markups. In particular, the output formatter watches for <...> and &...; sequences in the output stream, and wait to apply a caps or no-caps flag until after any markups have been passed. For example, if you display this:

"\^<a href='testing'><i>hello</i></a>\n";

then the effect of the caps flag on the output stream will be as follows:

<a href='testing'><i>Hello</i></a>\n

Note how the output formatter waits to apply the caps flag until after all of the intervening markups. This is desirable, since the caps flag would have effectively been "lost" if it had applied to the first alphabetic character after the flag - that would have been the "a" in the <a> tag, and there would have been no point in capitalizing that character, as it's just the name of a tag.

This behavior ensures that caps and nocaps flags won't be lost if the substituted parameter happens to contain some markup text. The flag will apply to the actual text displayed, not to HTML codes.

Notes for TADS 2 users

The display system in TADS 3 is essentially the same system that was used in TADS 2, but it has been substantially improved in a few areas. To summarize the differences from TADS 2: