1.4 The Printer
The Racket printer supports three modes:
write mode prints core datatypes in such a way that using read on the output produces a value that is equal? to the printed value;
display mode prints core datatypes in a more “end-user” style rather than “programmer” style; for example, a string displays as its content characters without surrounding "s or escapes;
print mode by default—
when print-as-expression is #t— prints most datatypes in such a way that evaluating the output as an expression produces a value that is equal? to the printed value; when print-as-expression is set to #f, then print mode is like write mode.
In print mode when print-as-expression is #t (as is the default), a value prints at a quoting depth of either 0 (unquoted) or 1 (quoted). The initial quoting depth is accepted as an optional argument by print, and printing of some compound datatypes adjusts the print depth for component values. For example, when a list is printed at quoting depth 0 and all of its elements are quotable, the list is printed with a ' prefix, and the list’s elements are printed at quoting depth 1.
When the print-graph parameter is set to #t, then the printer first scans an object to detect cycles. The scan traverses the components of pairs, mutable pairs, vectors, boxes (when print-box is #t), hash tables (when print-hash-table is #t), fields of structures exposed by struct->vector (when print-struct is #t), and fields of structures exposed by printing when the structure’s type has the prop:custom-write property. If print-graph is #t, then this information is used to print sharing through graph definitions and references (see Reading Graph Structure). If a cycle is detected in the initial scan, then print-graph is effectively set to #t automatically.
With the exception of displaying byte strings, printing is defined in terms of Unicode characters; see Ports for information on how a character stream is written to a port’s underlying byte stream.
1.4.1 Printing Symbols
Symbols containing spaces or special characters write using escaping \ and quoting |s. When the read-case-sensitive parameter is set to #f, then symbols containing uppercase characters also use escaping \ and quoting |s. In addition, symbols are quoted with |s or leading \ when they would otherwise print the same as a numerical constant or as a delimited . (when read-accept-dot is #t).
When read-accept-bar-quote is #t, |s are used in printing when one | at the beginning and one | at the end suffice to correctly print the symbol. Otherwise, \s are always used to escape special characters, instead of quoting them with |s.
When read-accept-bar-quote is #f, then | is not treated as a special character. The following are always special characters:
( ) [ ] } } " , ' ` ; \
In addition, # is a special character when it appears at the beginning of the symbol, and when it is not followed by %.
Symbols display without escaping or quoting special characters. That is, the display form of a symbol is the same as the display form of symbol->string applied to the symbol.
Symbols print the same as they write, unless print-as-expression is set to #t (as is the default) and the current quoting depth is 0. In that case, the symbol’s printed form is prefixed with '. For the purposes of printing enclosing datatypes, a symbol is quotable.
1.4.2 Printing Numbers
A number prints the same way in write, display, and print modes. For the purposes of printing enclosing datatypes, a symbol is quotable.
A complex number that is not a real number always prints as ‹m›+‹n›i, where ‹m› and ‹n› are the printed forms of its real and imaginary parts, respectively.
An inexact real number prints with either a . decimal point, an e exponent marker, or both. The form is selected so that the output is as short as possible, with the constraint that reading the printed form back in produces an equal? number.
An exact 0 prints as 0.
A positive, exact integer prints as a sequence of decimal digits that does not start with 0.
A positive, exact, real, non-integer number prints as ‹m›/‹n›, where ‹m› and ‹n› are the printed forms of the number’s numerator and denominator (as determined by numerator and denominator).
A negative exact number prints with a - prefix on the printed form of the number’s exact negation.
1.4.3 Printing Booleans
The boolean constant #t prints as #true or #t in all modes (display, write, and print), depending on the value of print-boolean-long-form, and the constant #f prints as #false or #f. For the purposes of printing enclosing datatypes, a symbol is quotable.
1.4.4 Printing Pairs and Lists
In write and display modes, an empty list prints as (). A pair normally prints starting with ( followed by the printed form of its car. The rest of the printed form depends on the cdr:
If the cdr is a pair or the empty list, then the printed form of the pair completes with the printed form of the cdr, except that the leading ( in the cdr’s printed form is omitted.
Otherwise, the printed for of the pair continues with a space, ., another space, the printed form of the cdr, and a ).
If print-reader-abbreviations is set to #t, then pair printing in write mode is adjusted in the case of a pair that starts a two-element list whose first element is 'quote, 'quasiquote, 'unquote, 'unquote-splicing, 'syntax, 'quasisyntax, 'unsyntax, or 'unsyntax-splicing. In that case, the pair is printed with the corresponding reader syntax: ', `, ,, ,@, #', #`, #,, or #,@, respectively. After the reader syntax, the second element of the list is printed. When the list is a tail of an enclosing list, the tail is printed after a . in the enclosing list (after which the reader abbreviations work), instead of including the tail as two elements of the enclosing list.
The printed form of a pair is the same in both write and display modes, except as the printed form of the pair’s car and cdr vary with the mode. The print form is also the same if print-as-expression is #f or the quoting depth is 1.
For print mode when print-as-expression is #t and the quoting depth is 0, then the empty list prints as '(). For a pair whose car and cdr are quotable, the pair prints in write mode but with a ' prefix; the pair’s content is printed with quoting depth 1. Otherwise, when the car or cdr is not quotable, then pair prints with either cons (when the cdr is not a pair), list (when the pair is a list), or list* (otherwise) after the opening (, any . that would otherwise be printed is suppressed, and the pair content is printed at quoting depth 0. In all cases, when print-as-expression is #t for print mode, then the value of print-reader-abbreviations is ignored and reader abbreviations are always used for lists printed at quoting depth 1.
By default, mutable pairs (as created with mcons) print the same as pairs for write and display, except that { and } are used instead of ( and ). Note that the reader treats {...} and (...) equivalently on input, creating immutable pairs in both cases. Mutable pairs in print mode with print-as-expression as #f or a quoting depth of 1 also use { and }. In print mode with print-as-expression as #t and a quoting depth of 0, a mutable pair prints as (mcons , the mcar and mcdr printed at quoting depth 0 and separated by a space, and a closing ).
If the print-pair-curly-braces parameter is set to #t, then pairs print using { and } when not using print mode with print-as-expression as #t and a quoting depth of 0. If the print-mpair-curly-braces parameter is set to #f, then mutable pairs print using ( and ) in that mode.
For the purposes of printing enclosing datatypes, an empty list is always quotable, a pair is quotable when its car and cdr are quotable, and a mutable list is never quotable.
1.4.5 Printing Strings
All strings display as their literal character sequences.
The write or print form of a string starts with " and ends with another ". Between the "s, each character is represented. Each graphic or blank character is represented as itself, with two exceptions: " is printed as \", and \ is printed as \\. Each non-graphic, non-blank character (according to char-graphic? and char-blank?) is printed using the escape sequences described in Reading Strings, using \a, \b, \t, \n, \v, \f, \r, or \e if possible, otherwise using \u with four hexadecimal digits or \U with eight hexadecimal digits (using the latter only if the character value does not fit into four digits).
All byte strings display as their literal byte sequence; this byte sequence may not be a valid UTF-8 encoding, so it may not correspond to a sequence of characters.
The write or print form of a byte string starts with #" and ends with a ". Between the "s, each byte is written using the corresponding ASCII decoding if the byte is between 0 and 127 and the character is graphic or blank (according to char-graphic? and char-blank?). Otherwise, the byte is written using \a, \b, \t, \n, \v, \f, \r, or \e if possible, otherwise using \ followed by one to three octal digits (only as many as necessary).
For the purposes of printing enclosing datatypes, a string or a byte string is quotable.
1.4.6 Printing Vectors
In display mode, the printed form of a vector is # followed by the printed form of vector->list applied to the vector. In write mode, the printed form is the same, except that when the print-vector-length parameter is #t, a decimal integer is printed after the #, and a repeated last element is printed only once.
Vectors print the same as they write, unless print-as-expression is set to #t and the current quoting depth is 0. In that case, if all of the vector’s elements are quotable, then the vector’s printed form is prefixed with ' and its elements printed with quoting depth 1. If its elements are not all quotable, then the vector prints as (vector , the elements at quoting depth 0, and a closing ). A vector is quotable when all of its elements are quotable.
1.4.7 Printing Structures
When the print-struct parameter is set to #t, then the way that structures print depends on details of the structure type for which the structure is an instance:
If the structure type is a prefab structure type, then it prints in write or display mode using #s( followed by the prefab structure type key, then the printed form of each field in the structure, and then ).
In print mode when print-as-expression is set to #t and the current quoting depth is 0, if the structure’s content is all quotable, then the structure’s printed form is prefixed with ' and its content is printed with quoting depth 1. If any of its content is not quotable, then the structure type prints the same as a non-prefab structure type.
An instance of a prefab structure type is quotable when all of its content is quotable.
If the structure has a prop:custom-write property value, then the associated procedure is used to print the structure, unless the print-unreadable parameter is set to #f.
For print mode, an instance of a structure type with a prop:custom-write property is treated as quotable if it has the prop:custom-print-quotable property with a value of 'always. If it has 'maybe as the property value, then the structure is treated as quotable if its content is quotable, where the content is determined by the values recursively printed by the structure’s prop:custom-write procedure. Finally, if the structure has 'self as the property value, then it is treated as quotable.
In print mode when print-as-expression is #t, the structure’s prop:custom-write procedure is called with either 0 or 1 as the quoting depth, normally depending on the structure’s prop:custom-print-quotable property value. If the property value is 'always, the quoting depth is normally 1. If the property value is 'maybe, then the quoting depth is 1 if the structure is quotable, or normally 0 otherwise. If the property value is 'self, then the quoting depth may be 0 or 1; it is normally 0 if the structure is not printed as a part of an enclosing quotable value, even though the structure is treated as quotable. Finally, if the property value is 'never, then the quoting depth is normally 0. The quoting depth can vary from its normal value if the structure is printed with an explicit quoting depth of 1.
If the structure’s type is transparent or if any ancestor is transparent (i.e., struct? on the instance produces #t), then the structure prints as the vector produced by struct->vector in display mode, in write mode, or in print mode when print-as-expression is set to #f or when the quoting depth is 0.
In print mode with print-as-expression as #t and a quoting depth of 0, the structure content is printed with a ( followed by the structure’s type name (as determined by object-name) in write mode; the remaining elements are printed at quoting depth 0 and separated by a space, and finally a closing ).
A transparent structure type that is not a prefab structure type is never quotable.
For any other structure type, the structure prints as an unreadable value; see Printing Unreadable Values for more information.
If the print-struct parameter is set to #f, then all structures without a prop:custom-write property print as unreadable values (see Printing Unreadable Values) and count as quotable.
1.4.8 Printing Hash Tables
When the print-hash-table parameter is set to #t, in write and display modes, a hash table prints starting with #hash(, #hasheqv(, or #hasheq( for a table using equal?, eqv?, or eq? key comparisons, respectively. After the prefix, each key–value mapping is shown as (, the printed form of a key, a space, ., a space, the printed form the corresponding value, and ), with an additional space if the key–value pair is not the last to be printed. After all key–value pairs, the printed form completes with ).
In print mode when print-as-expression is #f or the quoting depth is 1, the printed form is the same as for write. Otherwise, if the hash table’s keys and values are all quotable, the table prints with a ' prefix, and the table’s key and values are printed at quoting depth 1. If some key or value is not quotable, the hash table prints as (hash , (hasheqv , or (hasheq followed by alternating keys and values printed at quoting depth 1 and separated by spaces, and finally a closing ). A hash table is quotable when all of its keys and values are quotable.
When the print-hash-table parameter is set to #f, a hash table prints as #<hash> and counts as quotable.
1.4.9 Printing Boxes
When the print-box parameter is set to #t, a box prints as #& followed by the printed form of its content in write, display, or print mode when print-as-expression is #f or the quoting depth is 1.
In print mode when print-as-expression is #t and the quoting depth is 0, a box prints with a ' prefix and its value is printed at quoting depth 1 when its content is quotable, otherwise the box prints a (box followed by the content at quoting depth 0 and a closing ). A box is quotable when its content is quotable.
When the print-box parameter is set to #f, a box prints as #<box> and counts as quotable.
1.4.10 Printing Characters
Characters with the special names described in Reading Characters write and print using the same name. (Some characters have multiple names; the #\newline and #\nul names are used instead of #\linefeed and #\null.) Other graphic characters (according to char-graphic?) write as #\ followed by the single character, and all others characters are written in #\u notation with four digits or #\U notation with eight digits (using the latter only if the character value does not fit in four digits).
All characters display directly as themselves (i.e., a single character).
For the purposes of printing enclosing datatypes, a character is quotable.
1.4.11 Printing Keywords
Keywords write, print, and display the same as symbols (see Printing Symbols) except with a leading #: (after any ' prefix added in print mode), and without special handling for an initial # or when the printed form would match a number or a delimited . (since #: distinguishes the keyword).
For the purposes of printing enclosing datatypes, a keyword is quotable.
1.4.12 Printing Regular Expressions
Regexp values write, display, and print starting with #px (for pregexp-based regexps) or #rx (for regexp-based regexps) followed by the write form of the regexp’s source string or byte string.
For the purposes of printing enclosing datatypes, a regexp value is quotable.
1.4.13 Printing Paths
Paths write and print as #<path:....>. A path displays the same as the string produced by path->string. For the purposes of printing enclosing datatypes, a path counts as quotable.
Although a path can be converted to a string with path->string or to a byte string with path->bytes, neither is clearly the right choice for printing a path and reading it back. If the path value is meant to be moved among platforms, then a string is probably the right choice, despite the potential for losing information when converting a path to a string. For a path that is intended to be re-read on the same platform, a byte string is probably the right choice, since it preserves information in an unportable way. Paths do not print in a readable way so that programmers are not misled into thinking that either choice is always appropriate.
1.4.14 Printing Unreadable Values
For any value with no other printing specification, assuming that the print-unreadable parameter is set to #t, the output form is #<‹something›>, where ‹something› is specific to the type of the value and sometimes to the value itself. If print-unreadable is set to #f, then attempting to print an unreadable value raises exn:fail.
For the purposes of printing enclosing datatypes, a value that prints unreadably nevertheless counts as quotable.
1.4.15 Printing Compiled Code
Compiled code as produced by compile prints using #~. Compiled code printed with #~ is essentially assembly code for Racket, and reading such a form produces a compiled form when the read-accept-compiled parameter is set to #t.
When a compiled form contains syntax object constants, they must not be tainted or armed; the #~-marshaled form drops source-location information and properties (see Syntax Object Properties) for the syntax objects.
Compiled code parsed from #~ may contain references to unexported or protected bindings from a module. At read time, such references are associated with the current code inspector (see current-code-inspector), and the code will only execute if that inspector controls the relevant module invocation (see Code Inspectors).
A compiled-form object may contain uninterned symbols (see Symbols) that were created by gensym or string->uninterned-symbol. When the compiled object is read via #~, each uninterned symbol in the original form is mapped to a new uninterned symbol, where multiple instances of a single symbol are consistently mapped to the same new symbol. The original and new symbols have the same printed representation. Unreadable symbols, which are typically generated indirectly during expansion and compilation, are saved and restored consistently through #~.
The dynamic nature of uninterned symbols and their localization within #~ can cause problems when gensym or string->uninterned-symbol is used to construct an identifier for a top-level or module binding (depending on how the identifier and its references are compiled). To avoid problems, generate distinct identifiers either with generate-temporaries or by applying the result of make-syntax-introducer to an existing identifier; those functions lead to top-level and module variables with unreadable symbolic names, and the names are deterministic as long as expansion is otherwise deterministic.
Despite the problems inherent with uninterned symbols as variable names, they are partially supported even across multiple #~s: When compiled code contains a reference to a module-defined variable whose name is an uninterned symbol, the relative position of the variable among the module’s definitions is recorded, and the reference can be linked back to the definition based on its position and the characters in its name. This accommodation works only for variable references in compiled code; it does not work for syntax-quoted identifiers, for example.
Finally, a compiled form may contain path literals. Although paths are not normally printed in a way that can be read back in, path literals can be written and read as part of compiled code. The current-write-relative-directory parameter is used to convert the path to a relative path as is it written, and then current-load-relative-directory parameter is used to convert any relative path back as it is read. The relative-path conversion applies on reading whether the path was originally relative or not.