xlist

Fancy lists, with bounded or unbounded repetition of elements. Can be used as a type or match pattern.

To use the type expander, you must first require the type-expander library.

(xList Number ^ 3 Symbol String ^ 2)

(List Number Number Number Symbol String String)

The notation type * indicates that the given type may be repeated zero or more times. Therefore, the following two types are equivalent:

(xList Number * Symbol String *)

(Rec R1 (U (Pairof Number R1)

(List* Symbol (Rec R2 (U (Pairof String R2)

Null)))))

The notation type ^ n + indicates that the given type may be repeated n or more times. Therefore, the following two types are equivalent:

(xList Number ^ {2 +} String)

(List* Number Number (Rec R1 (U (Pairof Number R1)

(List String))))

When the number preceding + is omitted, it defaults to 1.

The notation type ^ once yields the same type as type ^ 1, but other forms recognise once and treat it specially. For example, xlist-split splits the corresponding element as a standalone value, not as a list of length one.

The notation type ^ n - m indicates that the given type may be repeated between n (inclusive) and m (inclusive) times. Therefore, the following two types are equivalent:

(xList Number ^ {2 - 5} String)

(U (List Number Number String)

(List Number Number Number String)

(List Number Number Number Number String)

(List Number Number Number Number Number String))

Be aware that the tail of the xList following the use of type ^ n - m is repeated n - m times, so if the tail itself contains uses of -, the resulting macro-expanded type will be huge, and may easily make Typed/Racket run out of memory, or slow down the type checking.

If the first bound is omitted, it defaults to 0, and if the second bound is omitted, it defaults to ∞. This means that - on its own is equivalent to *, but the latter form is preferred.

The superscripted-repeat is a representation of repeat using superscripted unicode characters, without spaces (i.e. the superscripted-repeat is a single identifier):

• Digits are replaced by their unicode superscripted counterparts ⁰¹²³⁴⁵⁶⁷⁸⁹

• + and - are replaced by their unicode superscripted counterparts, respectively ⁺ and ⁻

• * is replaced by the unicode character “COMBINING ASTERISK ABOVE”  ⃰ (code point U+20F0)

• ∞ is always omitted, as n - ∞ and - ∞ are equivalent to n - and 0 -

A superscripted-id is a type identifier ending with a sequence of characters which would otherwise be valid for superscripted-repeat. In other words, if the type is an identifier, the type and the superscripted-repeat can be coalesced into a single identifier.

The identifier String3 is equivalent to the notations String 3 (with a space between the identifier and the 3) and String ^ 3.

Similarly, the identifier String* is equivalent to the notations String * (with a space between the identifier and the *), String ^ * (using a regular asterisk, i.e. the multiplication function in Racket) and String * (using a regular asterisk, i.e. the multiplication function in Racket).

The same logic applies to the other cases.

If the repeat is once, or if the pattern does not have a repeat, then the pattern is not put under ellipses, so that (match '(42) [(xlist a ^ once) a]) returns 42, whereas (match '(42) [(xlist a ^ 1) a]) returns '(42).

Additionally, when #,@pattern appears as one of the xlist elements, the given pattern may match any number of elements in the list. This is implemented in terms of append from the match-string library.

The following two match patterns are therefore equivalent:

(xlist number?3 - 5 ,@(list-no-order number? string?) symbol?+)

(append (and (list number? ...) (app length (? (between/c 3 5))))

(list-no-order number? string?)

(list symbol? ..1))

Applying a repeat indicator on a splice is not supported yet, i.e. (xlist ,@(list-no-order number? string?) 5) will not work.

Note : Typed/Racket’s type inference is not strong enough (yet) to support some match patterns, and there is no typed/match library which would help with that (yet). This means that although by construction xlist tries to avoid generating such patterns, a few of the patterns supported by xlist will not work in typed/racket (rest values and spliced lists are the most likely to cause problems). As an alternative, try the split-xlist pattern, which produces code which should propagate type information to the different sub-lists.

1 Splitting an xlist in its constituent sublists

Note that split-xlist assumes the value it is matched against has the type (xlist τᵢ ... maybe-rest), but does not apply (? (make-predicate (xlist τᵢ ... maybe-rest))) to the value itself. The rationale is that the make-predicate may fail at compile-time if it cannot generate a contract for the given type. In some cases, however split-xlist will still manage to successfully generate the match pattern, and can be used on its own, provided that the value is statically known to be of the right type.

It is therefore recommended to use split-xlist as follows when the type of the value is not known to be acceptable by split-xlist:

2 Untyped versions of xlist

3 Special identifiers recognised by xlist