1 {-# LANGUAGE DeriveDataTypeable #-}
2 {-# LANGUAGE FlexibleInstances #-}
4 {-# LANGUAGE QuasiQuotes #-}
5 {-# LANGUAGE RecordWildCards #-}
6 {-# LANGUAGE StandaloneDeriving #-}
7 {-# LANGUAGE TemplateHaskell #-}
8 {-# LANGUAGE TypeFamilies #-}
10 -- | Parse TSN XML for the DTD "newsxml.dtd". Each document contains a
11 -- root element \<message\> that contains an entire news item.
17 -- * WARNING: these are private but exported to silence warnings
18 News_NewsLocationConstructor(..),
19 News_NewsTeamConstructor(..),
21 NewsLocationConstructor(..),
22 NewsTeamConstructor(..) )
26 import Data.Data ( Data, constrFields, dataTypeConstrs, dataTypeOf )
27 import Data.Time.Clock ( UTCTime )
28 import Data.List.Utils ( join, split )
29 import Data.Tuple.Curry ( uncurryN )
30 import Data.Typeable ( Typeable )
31 import Database.Groundhog (
34 import Database.Groundhog.Core ( DefaultKey )
35 import Database.Groundhog.TH (
39 import Test.Tasty ( TestTree, testGroup )
40 import Test.Tasty.HUnit ( (@?=), testCase )
41 import Text.XML.HXT.Core (
57 tsn_db_field_namer ) -- Used in a test
58 import TSN.DbImport ( DbImport(..), ImportResult(..), run_dbmigrate )
59 import TSN.Picklers ( xp_time_stamp )
60 import TSN.XmlImport ( XmlImport(..) )
61 import Xml ( FromXml(..), ToDb(..), pickle_unpickle, unpickleable )
65 -- | The database type for teams as they show up in the news.
68 NewsTeam { team_name :: String }
72 instance ToDb NewsTeam where
73 -- | The database representaion of a 'NewsTeam' is itself.
74 type Db NewsTeam = NewsTeam
76 -- | This is needed to define the XmlImport instance for NewsTeam.
78 instance FromXml NewsTeam where
79 -- | How to we get a 'NewsTeam' from itself?
82 -- | Allow us to call 'insert_xml' on the XML representation of
85 instance XmlImport NewsTeam
88 -- | Mapping between News records and NewsTeam records in the
89 -- database. We don't name the fields because we don't use the names
90 -- explicitly; that means we have to give them nice database names
93 data News_NewsTeam = News_NewsTeam
98 -- | The database type for locations as they show up in the news.
102 city :: Maybe String,
103 state :: Maybe String,
107 instance ToDb NewsLocation where
108 -- | The database representation of a 'NewsLocation' is itself.
109 type Db NewsLocation = NewsLocation
111 -- | This is needed to define the XmlImport instance for NewsLocation.
113 instance FromXml NewsLocation where
114 -- | How to we get a 'NewsLocation' from itself?
117 -- | Allow us to call 'insert_xml' on the XML representation of
120 instance XmlImport NewsLocation
123 -- | Mapping between News records and NewsLocation records in the
124 -- database. We don't name the fields because we don't use the names
125 -- explicitly; that means we have to give them nice database names
128 data News_NewsLocation = News_NewsLocation
130 (DefaultKey NewsLocation)
133 -- | The msg_id child of <message> contains an event_id attribute; we
134 -- embed it into the 'News' type. We (pointlessly) use the "db_"
135 -- prefix here so that the two names don't collide on "id" when
136 -- Groundhog is creating its fields using our field namer.
140 db_event_id :: Maybe Int }
141 deriving (Data, Eq, Show, Typeable)
144 -- | The XML representation of a news item (message).
148 xml_xml_file_id :: Int,
149 xml_heading :: String,
151 xml_category :: String,
153 xml_url :: Maybe String,
154 xml_teams :: [NewsTeam],
155 xml_locations :: [NewsLocation],
157 xml_editor :: Maybe String,
158 xml_text :: Maybe String, -- Text and continue seem to show up in pairs,
159 xml_continue :: Maybe String, -- either both present or both missing.
160 xml_time_stamp :: UTCTime }
164 -- | The database representation of a news item. We drop several
165 -- uninteresting fields from 'Message', and omit the list fields which
166 -- will be represented as join tables.
172 db_url :: Maybe String,
174 db_editor :: Maybe String,
175 db_text :: Maybe String,
176 db_continue :: Maybe String,
177 db_time_stamp :: UTCTime }
178 deriving (Data, Eq, Show, Typeable)
182 instance ToDb Message where
183 type Db Message = News
185 -- | Convert the XML representation 'Message' to the database
186 -- representation 'News'.
188 instance FromXml Message where
189 -- | We use a record wildcard so GHC doesn't complain that we never
190 -- used the field names.
192 from_xml Message{..} = News { db_mid = xml_mid,
193 db_sport = xml_sport,
196 db_editor = xml_editor,
198 db_continue = xml_continue,
199 db_time_stamp = xml_time_stamp }
201 -- | This lets us call 'insert_xml' on a 'Message'.
203 instance XmlImport Message
205 -- | Define 'dbmigrate' and 'dbimport' for 'Message's. The import is
206 -- slightly non-generic because of our 'News_NewsTeam' and
207 -- 'News_NewsLocation' join tables.
209 instance DbImport Message where
212 migrate (undefined :: NewsTeam)
213 migrate (undefined :: NewsLocation)
214 migrate (undefined :: News)
215 migrate (undefined :: News_NewsTeam)
216 migrate (undefined :: News_NewsLocation)
218 dbimport message = do
219 -- Insert the message and acquire its primary key (unique ID)
220 news_id <- insert_xml message
222 -- And insert each one into its own table. We use insert_xml_or_select
223 -- because we know that most teams will already exist, and we
224 -- want to get back the id for the existing team when
225 -- there's a collision.
226 nt_ids <- mapM insert_xml_or_select (xml_teams message)
228 -- Now that the teams have been inserted, create
229 -- news__news_team records mapping beween the two.
230 let news_news_teams = map (News_NewsTeam news_id) nt_ids
231 mapM_ insert_ news_news_teams
233 -- Do all of that over again for the NewsLocations.
234 loc_ids <- mapM insert_xml_or_select (xml_locations message)
235 let news_news_locations = map (News_NewsLocation news_id) loc_ids
236 mapM_ insert_ news_news_locations
238 return ImportSucceeded
241 -- These types don't have special XML representations or field name
242 -- collisions so we use the defaultCodegenConfig and give their
243 -- fields nice simple names.
244 mkPersist defaultCodegenConfig [groundhog|
250 - name: unique_news_team
254 - entity: NewsLocation
255 dbName: news_locations
259 - name: unique_news_location
261 fields: [city, state, country]
266 -- These types have fields with e.g. db_ and xml_ prefixes, so we
267 -- use our own codegen to peel those off before naming the columns.
268 mkPersist tsn_codegen_config [groundhog|
276 - {name: msg_id, dbName: msg_id}
277 - {name: event_id, dbName: event_id}
286 - entity: News_NewsTeam
287 dbName: news__news_teams
289 - name: News_NewsTeam
291 - name: news_NewsTeam0 # Default created by mkNormalFieldName
295 - name: news_NewsTeam1 # Default created by mkNormalFieldName
296 dbName: news_teams_id
300 - entity: News_NewsLocation
301 dbName: news__news_locations
303 - name: News_NewsLocation
305 - name: news_NewsLocation0 # Default created by mkNormalFieldName
309 - name: news_NewsLocation1 # Default created by mkNormalFieldName
310 dbName: news_locations_id
316 -- | Convert a 'NewsTeam' to/from XML.
318 pickle_news_team :: PU NewsTeam
321 xpWrap (from_string, to_string) xpText
323 to_string :: NewsTeam -> String
324 to_string = team_name
326 from_string :: String -> NewsTeam
327 from_string = NewsTeam
330 -- | Convert a 'MsgId' to/from XML.
332 pickle_msg_id :: PU MsgId
335 xpWrap (from_tuple, to_tuple) $
336 xpPair xpInt (xpAttr "EventId" (xpOption xpInt))
338 from_tuple = uncurryN MsgId
339 to_tuple m = (db_msg_id m, db_event_id m)
342 -- | Convert a 'NewsLocation' to/from XML.
344 pickle_location :: PU NewsLocation
347 xpWrap (from_tuple, to_tuple) $
348 xpTriple (xpOption (xpElem "city" xpText))
349 (xpOption (xpElem "state" xpText))
350 (xpElem "country" xpText)
353 uncurryN NewsLocation
354 to_tuple l = (city l, state l, country l)
357 -- | Convert a 'Message' to/from XML.
359 pickle_message :: PU Message
362 xpWrap (from_tuple, to_tuple) $
363 xp13Tuple (xpElem "XML_File_ID" xpInt)
364 (xpElem "heading" xpText)
366 (xpElem "category" xpText)
367 (xpElem "sport" xpText)
368 (xpElem "url" $ xpOption xpText)
369 (xpList pickle_news_team)
370 (xpList pickle_location)
371 (xpElem "SMS" xpText)
372 (xpOption (xpElem "Editor" xpText))
373 (xpOption (xpElem "text" xpText))
375 (xpElem "time_stamp" xp_time_stamp)
377 from_tuple = uncurryN Message
378 to_tuple m = (xml_xml_file_id m, -- Verbose,
379 xml_heading m, -- but
380 xml_mid m, -- eliminates
381 xml_category m, -- GHC
382 xml_sport m, -- warnings
385 xml_locations m, -- .
392 -- | We combine all of the \<continue\> elements into one 'String'
393 -- while unpickling and do the reverse while pickling.
395 pickle_continue :: PU (Maybe String)
398 xpWrap (to_string, from_string) $
400 xpList (xpElem "P" xpText)
402 from_string :: String -> [String]
403 from_string = split "\n"
405 to_string :: [String] -> String
406 to_string = join "\n"
413 -- | A list of all tests for this module.
415 news_tests :: TestTree
419 [ test_news_fields_have_correct_names,
420 test_pickle_of_unpickle_is_identity,
421 test_unpickle_succeeds ]
424 -- | Make sure our codegen is producing the correct database names.
426 test_news_fields_have_correct_names :: TestTree
427 test_news_fields_have_correct_names =
428 testCase "news fields get correct database names" $
429 mapM_ check (zip actual expected)
431 -- This is cool, it uses the (derived) Data instance of
432 -- News.News to get its constructor names.
433 field_names :: [String]
435 constrFields . head $ dataTypeConstrs $ dataTypeOf (undefined :: News)
439 map (\x -> tsn_db_field_namer "herp" "derp" 8675309 x 90210) field_names
442 actual = ["mid", "sport", "url", "sms", "editor", "text", "continue"]
444 check (x,y) = (x @?= y)
447 -- | If we unpickle something and then pickle it, we should wind up
448 -- with the same thing we started with. WARNING: success of this
449 -- test does not mean that unpickling succeeded.
451 test_pickle_of_unpickle_is_identity :: TestTree
452 test_pickle_of_unpickle_is_identity = testGroup "pickle-unpickle tests"
453 [ check "pickle composed with unpickle is the identity"
454 "test/xml/newsxml.xml",
456 check "pickle composed with unpickle is the identity (with Editor)"
457 "test/xml/newsxml-with-editor.xml" ]
459 check desc path = testCase desc $ do
460 (expected, actual) <- pickle_unpickle pickle_message path
464 -- | Make sure we can actually unpickle these things.
466 test_unpickle_succeeds :: TestTree
467 test_unpickle_succeeds = testGroup "unpickle tests"
468 [ check "unpickling succeeds"
469 "test/xml/newsxml.xml",
471 check "unpickling succeeds (with Editor)"
472 "test/xml/newsxml-with-editor.xml" ]
474 check desc path = testCase desc $ do
475 actual <- unpickleable path pickle_message