X-Git-Url: http://gitweb.michael.orlitzky.com/?a=blobdiff_plain;f=src%2FXml.hs;h=95eddc0e51f949acbb287003dfb98c22ff244cb4;hb=5124833f973f6c47a4eda3f4e343d3d8631897c6;hp=129c6e9cebe4236b7af99a7988c168b662eefb44;hpb=988f693ce7f1abb6566e75d539ac312b627c31d5;p=dead%2Fhtsn-import.git diff --git a/src/Xml.hs b/src/Xml.hs index 129c6e9..95eddc0 100644 --- a/src/Xml.hs +++ b/src/Xml.hs @@ -1,40 +1,58 @@ +{-# LANGUAGE TypeFamilies #-} + -- | General XML stuff. -- module Xml ( + DtdName(..), + FromXml(..), parse_opts, - pickle_unpickle ) + pickle_unpickle, + unpickleable ) where +import Control.Exception ( SomeException(..), catch ) import Text.XML.HXT.Core ( (>>>), (/>), + PU, SysConfigList, XmlPickler(..), hasName, - no, readDocument, runX, - withPreserveComment, withRemoveWS, - withSubstDTDEntities, - withValidate, xpickleVal, xunpickleVal, yes ) + +-- | A typeclass for XML types that can be converted into an associated +-- database type. The story behind this is long, but basically, we need +-- to different types for each XML thingie we're going to import: a +-- database type and an XML type. Both Groundhog and HXT are very +-- particular about the types that they can use, and there's no way +-- to reuse e.g. a type that HXT can pickle in Groundhog. So this +-- typeclass gives us a way to get the database type from the XML +-- type that we have to define for HXT. +-- +class FromXml a where + -- | Each instance a must declare its associated database type (Db a) + type Db a :: * + + -- | And provide a function for getting a (Db a) out of an "a". + from_xml :: a -> Db a + + +-- | Represents the DTD filename ("SYSTEM") part of the DOCTYPE +-- definition. +newtype DtdName = DtdName String + -- | A list of options passed to 'readDocument' when we parse an XML --- document. We don't validate because the DTDs from TSN are --- wrong. As a result, we don't want to keep useless DTDs --- areound. Thus we disable 'withSubstDTDEntities' which, when --- combined with "withValidate no", prevents HXT from trying to read --- the DTD at all. +-- document. All cosmetic whitespace should be removed, otherwise we +-- have to parse it in each pickler. -- parse_opts :: SysConfigList -parse_opts = - [ withPreserveComment no, - withRemoveWS yes, - withSubstDTDEntities no, - withValidate no ] +parse_opts = [ withRemoveWS yes ] -- | Given a root element name and a file path, return both the @@ -48,6 +66,11 @@ parse_opts = -- to 'xpickle' would be ambiguous. By returning some @a@s, we allow -- the caller to annotate its type. -- +-- Note that this will happily pickle nothing to nothing and then +-- unpickle it back to more nothing. So the fact that the +-- before/after results from this function agree does not mean that +-- the document was successfully unpickled! +-- pickle_unpickle :: XmlPickler a => String -> FilePath @@ -55,7 +78,7 @@ pickle_unpickle :: XmlPickler a pickle_unpickle root_element filepath = do -- We need to check only the root message element since -- readDocument produces a bunch of other junk. - expected <- runX $ arr_getobj + expected <- runX arr_getobj actual <- runX $ arr_getobj >>> xpickleVal xpickle @@ -69,3 +92,20 @@ pickle_unpickle root_element filepath = do hasName root_element >>> xunpickleVal xpickle + + + +-- | Is the given XML file unpickleable? Unpickling will be attempted +-- using the @unpickler@ argument. If we unilaterally used the +-- generic 'xpickle' function for our unpickler, a type ambiguity +-- would result. By taking the unpickler as an argument, we allow +-- the caller to indirectly specify a concrete type. +-- +unpickleable :: XmlPickler a => FilePath -> PU a -> IO Bool +unpickleable filepath unpickler = do + xmldoc <- try_unpickle `catch` (\(SomeException _) -> return []) + return $ (not . null) xmldoc + where + try_unpickle = runX $ readDocument parse_opts filepath + >>> + xunpickleVal unpickler