Text.XML.HaXml.Combinators

This module defines the notion of filters and filter combinators for processing XML documents.

These XML transformation combinators are described in the paper ``Haskell and XML: Generic Combinators or Type-Based Translation?'' Malcolm Wallace and Colin Runciman, Proceedings ICFP'99.

The content filter type.

type CFilter i = Content i -> [Content i]

All document transformations are content filters. A filter takes a single XML Content value and returns a sequence of Content values, possibly empty.

Simple filters.

Selection filters.

In the algebra of combinators, none is the zero, and keep the identity. (They have a more general type than just CFilter.)

keep :: a -> [a]

none :: a -> [b]

children :: CFilter i

Throw away current node, keep just the (unprocessed) children.

position :: Int -> CFilter i -> CFilter i

Select the n'th positional result of a filter.

Predicate filters.

These filters either keep or throw away some content based on a simple test. For instance, elm keeps only a tagged element, txt keeps only non-element text, tag keeps only an element with the named tag, attr keeps only an element with the named attribute, attrval keeps only an element with the given attribute value, tagWith keeps only an element whose tag name satisfies the given predicate.

elm :: CFilter i

txt :: CFilter i

tag :: String -> CFilter i

attr :: Name -> CFilter i

attrval :: Attribute -> CFilter i

tagWith :: (String -> Bool) -> CFilter i

Search filters.

find :: String -> (String -> CFilter i) -> CFilter i

For a mandatory attribute field, find key cont looks up the value of the attribute name key, and applies the continuation cont to the value.

iffind :: String -> (String -> CFilter i) -> CFilter i -> CFilter i

When an attribute field may be absent, use iffind key yes no to lookup its value. If the attribute is absent, it acts as the no filter, otherwise it applies the yes filter.

ifTxt :: (String -> CFilter i) -> CFilter i -> CFilter i

ifTxt yes no processes any textual content with the yes filter, but otherwise is the same as the no filter.

Filter combinators

Basic combinators.

o :: CFilter i -> CFilter i -> CFilter i

Sequential (Irish,backwards) composition

union :: (a -> [b]) -> (a -> [b]) -> a -> [b]

Binary parallel composition. Each filter uses a copy of the input, rather than one filter using the result of the other. (Has a more general type than just CFilter.)

cat :: [a -> [b]] -> a -> [b]

Glue a list of filters together. (A list version of union; also has a more general type than just CFilter.)

andThen :: (a -> c) -> (c -> a -> b) -> a -> b

A special form of filter composition where the second filter works over the same data as the first, but also uses the first's result.

(|>|) :: (a -> [b]) -> (a -> [b]) -> a -> [b]

Directional choice: in f |>| g give g-productions only if no f-productions

with :: CFilter i -> CFilter i -> CFilter i

Pruning: in f with g, keep only those f-productions which have at least one g-production

without :: CFilter i -> CFilter i -> CFilter i

Pruning: in f without g, keep only those f-productions which have no g-productions

(/>) :: CFilter i -> CFilter i -> CFilter i

Pronounced slash, f /> g means g inside f

(</) :: CFilter i -> CFilter i -> CFilter i

Pronounced outside, f </ g means f containing g

et :: (String -> CFilter i) -> CFilter i -> CFilter i

Join an element-matching filter with a text-only filter

path :: [CFilter i] -> CFilter i

Express a list of filters like an XPath query, e.g. path [children, tag "name1", attr "attr1", children, tag "name2"] is like the XPath query /name1[@attr1]/name2.

Recursive search.

Recursive search has three variants: deep does a breadth-first search of the tree, deepest does a depth-first search, multi returns content at all tree-levels, even those strictly contained within results that have already been returned.

deep :: CFilter i -> CFilter i

deepest :: CFilter i -> CFilter i

multi :: CFilter i -> CFilter i

Interior editing.

when :: CFilter i -> CFilter i -> CFilter i

Interior editing: f when g applies f only when the predicate g succeeds, otherwise the content is unchanged.

guards :: CFilter i -> CFilter i -> CFilter i

Interior editing: g guards f applies f only when the predicate g succeeds, otherwise the content is discarded.

chip :: CFilter i -> CFilter i

Process CHildren In Place. The filter is applied to any children of an element content, and the element rebuilt around the results.

inplace :: CFilter i -> CFilter i

Process an element In Place. The filter is applied to the element itself, and then the original element rebuilt around the results.

foldXml :: CFilter i -> CFilter i

Recursive application of filters: a fold-like operator. Defined as f o chip (foldXml f).

Constructive filters.

mkElem :: String -> [CFilter i] -> CFilter i

Build an element with the given tag name - its content is the results of the given list of filters.

mkElemAttr :: String -> [(String, CFilter i)] -> [CFilter i] -> CFilter i

Build an element with the given name, attributes, and content.

literal :: String -> CFilter i

Build some textual content.

cdata :: String -> CFilter i

Build some CDATA content.

replaceTag :: String -> CFilter i

Rename an element tag (leaving attributes in place).

replaceAttrs :: [(String, String)] -> CFilter i

Replace the attributes of an element (leaving tag the same).

C-like conditionals.

These definitions provide C-like conditionals, lifted to the filter level.

The (cond ? yes : no) style in C becomes (cond ?> yes :> no) in Haskell.

data ThenElse a

Conjoin the two branches of a conditional.

Constructors

a :> a

(?>) :: (a -> [b]) -> ThenElse (a -> [b]) -> a -> [b]

Select between the two branches of a joined conditional.

Filters with labelled results.

type LabelFilter i a = Content i -> [(a, Content i)]

A LabelFilter is like a CFilter except that it pairs up a polymorphic value (label) with each of its results.

Using and combining labelled filters.

oo :: (a -> CFilter i) -> LabelFilter i a -> CFilter i

Compose a label-processing filter with a label-generating filter.

x :: (CFilter i -> LabelFilter i a) -> (CFilter i -> LabelFilter i b) -> CFilter i -> LabelFilter i (a, b)

Combine labels. Think of this as a pair-wise zip on labels. e.g. (numbered x tagged)

Some label-generating filters.

numbered :: CFilter i -> LabelFilter i Int

Number the results from 1 upwards.

interspersed :: String -> CFilter i -> String -> LabelFilter i String

In interspersed a f b, label each result of f with the string a, except for the last one which is labelled with the string b.

tagged :: CFilter i -> LabelFilter i String

Label each element in the result with its tag name. Non-element results get an empty string label.

attributed :: String -> CFilter i -> LabelFilter i String

Label each element in the result with the value of the named attribute. Elements without the attribute, and non-element results, get an empty string label.

textlabelled :: CFilter i -> LabelFilter i (Maybe String)

Label each textual part of the result with its text. Element results get an empty string label.

extracted :: (Content i -> a) -> CFilter i -> LabelFilter i a

Label each content with some information extracted from itself.