Underscore

As early output devices (Teleprinters, CRTs and line printers) could not produce more than one character at a location, it was not possible to underscore text, so early encodings such as ITA2 and the first versions of ASCII had no underscore. IBM's EBCDIC character-coding system, introduced in 1964, added the underscore, which IBM referred to as the "break character". IBM's report on NPL (the early name of what is now called PL/I) leaves the character set undefined, but specifically mentions the break character, and gives RATE_OF_PAY as an example identifier.^[8] By 1967 the underscore had spread to ASCII,^[9] replacing the similarly shaped left-arrow character, ← (see also: PIP). C, developed at Bell Labs in the early 1970s, allowed the underscore in identifiers.^[10]

Underscore predates the existence of lower-case letters in many systems, so often it had to be used to make multi-word identifiers, since camelCase (see below) was not available.

Underscores inserted between letters are very common to make a "multi-word" identifier in languages that cannot handle spaces in identifiers. This convention is known as "snake_case" (the other popular method is called camelCase, where capital letters are used to show where the words start).

An underscore as the first character in an ID is often used to indicate an internal implementation that is not considered part of the API and should not be called by code outside that implementation. In Dart, all private properties of classes must start with an underscore; this usage is also common in other languages such as C++ even though those provide keywords to indicate that members are private. It is extensively used to hide variables and functions used for implementations in header files. In fact, the use of a single underscore for this became so common that C compilers had to standardize on a double leading underscore (for instance __DATE__) for actual built-in variables to avoid conflicts with the ones in header files. PHP "reserves all function names starting with __ [double underscore] as magical."^[11]

Python uses names that both start and end with double underscores (so called "dunder methods", as in double underscore) for magic members used for purposes such as operator overloading and reflection, and names starting but not ending with a double underscore to denote private member variables of classes which should be mangled in a manner which prevents them from colliding with members of derived classes unless the classes have the same name (__bar in class Foo will be mangled to _Foo__bar). By convention, members starting with a single underscore are considered private or protected, although this behavior only has inherent effect for modules, where import * statements by default import all names that do not start with an underscore, unless an export list is explicitly defined by the module.

A variable named with just an underscore often has special meaning. $_ or _ is the previous command or result in many interactive shells, such as those of Python, Ruby, and Perl. In Perl, @_ is a special array variable that holds the arguments to a function. In Clojure, it indicates an argument whose value will be ignored.^[12] In Python, it is sometimes used as a dummy variable, when its value is not important but the syntax requires it.^[13] In C#, it is used for discards, placeholder variables that are required by syntax but remain unassigned and not processed.^[14]

In some languages with pattern matching, such as Haskell and other ML-family languages, Rust, C#, Python, Prolog, and the Wolfram Language, the pattern _ matches any value, but does not perform binding, hence it can be reused multiple times in a single pattern and does not require declaration.

The ASCII underscore character can be inserted with the entities_ or _ (or _ or _).

HTML has a presentational element<u> that was originally used to underline text; this usage was deprecated in HTML4 in favor of the CSS style {text-decoration: underline}.^[15] In HTML5, the tag reappeared but its meaning was changed significantly: it now "represents a span of inline text which should be rendered in a way that indicates that it has a non-textual annotation".^[15] This facility is intended for example to provide a red wavy line (or wiggly line) underline to flag spelling errors at input time but which are not to be embedded in any stored file (unlike an emphasis mark, which would be). Other styles are also available: doubled, dotted, and dashed.^[16]

The elements may also exist in other markup languages, such as MediaWiki. The Text Encoding Initiative (TEI) provides an extensive selection of related elements for marking editorial activity (insertion, deletion, correction, addition, etc.).

Unicode has a free-standing underscore _ at U+005F, inherited from ASCII, which is a legacy of the typewriter practice of underlining using backspace and overtype. Modern practice uses the combining diacriticU+0332◌̲COMBINING LOW LINE that results in an underline when run together: u̲n̲d̲e̲r̲l̲i̲n̲e̲. Unicode also has U+0333◌̳COMBINING DOUBLE LOW LINE. In addition, there are single line and double line versions of the combining macron below, a diacritic that applies to single letters only.^[3]

In plain-text applications, including plain-text e-mails where emphasis markup is not possible, the desired emphasis is often indicated by surrounding words with underscore characters. For example, "You must use _emulsion_ paint on the ceiling".

Some applications will automatically add emphasis to text manually bracketed by underscores, either by underlining or by italicizing it (e.g. _string_ may render as either string or string).

Underline (typically red or wavy or both) is often used by spell checkers (and grammar checkers) to denote misspelled or otherwise incorrect text.