.PAS File Format – Lexical structure

Tokens

The Free Pascal Reference guide names five kinds of token: reserved words, identifiers, operators, separators (white space) and constants (numbers and strings). Names and reserved words are not case sensitive: BEGIN, Begin and begin are the same word, and so are MyProc and myproc.

Comments

Free Pascal lists (* and *) as equivalent to { and }. A comment ends with the closing delimiter of the form that opened it, except in ISO mode ({$mode iso} and Extended Pascal mode), where the standard lets a comment opened with { end with *) and one opened with (* end with }.

Nesting

• A comment of one kind may hold the other two kinds as plain text: { (* x *) }, (* { x } *) and { // x } are each one comment.
• A comment of the same kind nests only where the mode allows it. In Free Pascal’s own mode (fpc, the default) and in objfpc mode the compiler counts the levels, so { a { b } c } is one comment (the mode switch nestedcomments turns this on or off). In delphi, tp, macpas and iso modes, and in Delphi and Turbo Pascal themselves, a comment ends at the first closing delimiter, so { a { b } is a whole comment and c } is code.
• A // comment that holds a { or (* whose closing delimiter is on a later line is an error in Free Pascal.

The file viewer on this site follows the mode directives it meets ({$mode objfpc} and {$mode fpc} turn nesting on, {$mode delphi}, tp, macpas turn it off, iso and extendedpascal let either closing delimiter end a comment, and {$modeswitch nestedcomments} switches it). A file with no mode directive is read as Delphi or Turbo Pascal, without nesting.

Identifiers

• An identifier starts with a letter or an underscore and goes on with letters, digits and underscores. ISO 7185 allows no underscore, only letters and digits; Turbo Pascal, Delphi and Free Pascal allow it. Free Pascal counts between 1 and 127 significant characters.
• Delphi and Free Pascal also accept letters above code 127 (Unicode letters) in names.
• A reserved word may be used as a name by writing & in front of it: procedure &do; declares a procedure called do, and &do; calls it (Delphi, and Free Pascal 2.5.1 and later in its Free Pascal and Delphi modes).
• A name may be qualified by its unit (SysUtils.IntToStr), its record or object (P.X) or, in Delphi and Free Pascal, a unit scope (System.SysUtils.IntToStr).
• The directives (below) are not reserved. They are ordinary names except where they are used as directives, so name, index, read and write can be variables.

Reserved words

A reserved word cannot be redefined. Each dialect has more than the one before.

In Free Pascal’s tp mode only the Turbo Pascal words are reserved, and the Delphi ones can be redefined.

Directives and modifiers

These words are meaningful only where a directive may stand: after a procedure, function or method heading, in a property declaration, in a class declaration after class, and in a few other places. Elsewhere they are plain names.

Free Pascal’s Reference guide lists these modifiers: absolute abstract alias assembler bitpacked break cdecl continue cppdecl cvar default deprecated dynamic enumerator experimental export external far far16 forward generic helper implements index interrupt iocheck local message name near nodefault noreturn nostackframe oldfpccall otherwise overload override pascal platform private protected public published read register reintroduce result safecall saveregisters softfloat specialize static stdcall stored strict unaligned unimplemented varargs virtual winapi write. Delphi also has at contains delayed package readonly requires resident sealed unsafe writeonly (from memory; DocWiki was not read).

Symbols and operators

The characters with a special meaning in Free Pascal are + - * / = < > [ ] . , ( ) : ^ @ { } $ # & % and the pairs << >> ** <> >< <= >= := += -= *= /= (* *) (. .) //. In a range specifier, (. and .) stand for [ and ] (the alternative tokens of ISO 7185, for keyboards without brackets). These pairs keep their ordinary meaning inside strings.

The precedence of the operators is on the Expressions page.

Numbers

Hexadecimal literals have no h suffix and no 0x prefix in Pascal (the inline assembler does accept 0FFh and 0xFF). There are no suffixes for the type of a number and no literal for a real in hexadecimal.

Character and string literals

• Quoted string. Zero or more characters between single quotes, on one line. Two quotes in a row stand for one quote: 'it''s'. '' is the empty string, and a one-character string is also a character constant. There are no backslash escapes. A quoted string cannot hold a line end.
• Character code. # followed by a number gives the character with that code: #27, #$1B. Free Pascal also reads #&33 (octal) and #%11011 (binary). Codes above 255 give a wide character, and above 65535 a surrogate pair; the Free Pascal compiler accepts codes up to #$10FFFF.
• Control character. ^ and a letter: ^A is code 1, ^M is 13, ^Z is 26. ^@ is 0, ^[ is 27, and the characters @ A–Z [ \ ] ^ _ give 0 to 31. The caret is a control character only where a pointer type or a dereference cannot stand, so p^ and ^Integer are pointer syntax.
• Concatenation. Quoted strings and character codes may be written next to each other without an operator and form one string: 'Line one'#13#10'Line two', 'x'^G'y'. That is the only way to put a control character in a string.
• Multi-line string. Delphi 12 and the development version of Free Pascal (3.3.1, behind the multilinestrings mode switch, which its Object Pascal and Delphi modes turn on) take a string that opens with ''' (three quotes, or any odd number of at least three) as the last thing on a line, and closes with a line that holds only white space and the same number of quotes. The text in between, line ends included, is the string; the indentation of the closing line is taken off every line. ''' followed by text on the same line is not a multi-line string.
• Backtick string. The Free Pascal compiler source also reads `…` as a multi-line string (a doubled backtick is one backtick), behind the same mode switch.
• Double quotes. Oxygene also writes strings with double quotes. Standard Pascal, Delphi and Free Pascal do not.
• Encoding. A string is 8-bit characters (ASCII or UTF-8) or 16-bit wide characters. In Free Pascal source the characters are read in the code page of the file, UTF-8 by default; Delphi 2009 and later read UTF-8 or UTF-16 files.

WriteLn('It''s ', 'a ''quoted'' word');
S := 'tab:'#9'here'#13#10;
Esc := #27 + '[0m';
Bell := ^G;
Hex := #$41#$42;
Text := '''
  First line
    Indented line
  ''';

Labels

A label in standard Pascal is an unsigned integer of up to four digits (0 to 9999). Turbo Pascal, Delphi and Free Pascal also allow identifiers as labels. Labels are declared with label and used with goto (see Statements). In Free Pascal the {$GOTO} directive controls whether goto and label are allowed.

How the viewer colours it