X-Git-Url: http://gitweb.michael.orlitzky.com/?a=blobdiff_plain;f=src%2FTSN%2FXML%2FNews.hs;h=10c51956fa20c23a16ff45b3469bc477aac078a7;hb=9b28d5ed0e1570f0890eae1ee01721688c20e266;hp=6e10aed9d166089d7119b543066340a28596f717;hpb=d476419274cdf2997d768444cecc47922a902fdf;p=dead%2Fhtsn-import.git diff --git a/src/TSN/XML/News.hs b/src/TSN/XML/News.hs index 6e10aed..10c5195 100644 --- a/src/TSN/XML/News.hs +++ b/src/TSN/XML/News.hs @@ -3,34 +3,41 @@ {-# LANGUAGE GADTs #-} {-# LANGUAGE QuasiQuotes #-} {-# LANGUAGE RecordWildCards #-} -{-# LANGUAGE StandaloneDeriving #-} {-# LANGUAGE TemplateHaskell #-} {-# LANGUAGE TypeFamilies #-} --- | Parse TSN XML for the DTD "newsxml.dtd". Each document contains a --- root element \ that contains an entire news item. +-- | Parse TSN XML for the DTD \"newsxml.dtd\". Each document contains +-- a root element \ that contains an entire news item. -- module TSN.XML.News ( + dtd, + has_only_single_sms, pickle_message, -- * Tests news_tests, -- * WARNING: these are private but exported to silence warnings - News_NewsLocationConstructor(..), + News_LocationConstructor(..), News_NewsTeamConstructor(..), NewsConstructor(..), - NewsLocationConstructor(..), NewsTeamConstructor(..) ) where -- System imports. import Data.Data ( Data, constrFields, dataTypeConstrs, dataTypeOf ) +import Data.Time.Clock ( UTCTime ) import Data.List.Utils ( join, split ) import Data.Tuple.Curry ( uncurryN ) import Data.Typeable ( Typeable ) import Database.Groundhog ( + countAll, + deleteAll, insert_, - migrate ) + migrate, + runMigration, + silentMigrationLogger ) import Database.Groundhog.Core ( DefaultKey ) +import Database.Groundhog.Generic ( runDbConn ) +import Database.Groundhog.Sqlite ( withSqliteConn ) import Database.Groundhog.TH ( defaultCodegenConfig, groundhog, @@ -39,6 +46,16 @@ import Test.Tasty ( TestTree, testGroup ) import Test.Tasty.HUnit ( (@?=), testCase ) import Text.XML.HXT.Core ( PU, + XmlTree, + (/>), + (>>>), + addNav, + descendantAxis, + filterAxis, + followingSiblingAxis, + hasName, + remNav, + runLA, xp13Tuple, xpAttr, xpElem, @@ -47,86 +64,43 @@ import Text.XML.HXT.Core ( xpOption, xpPair, xpText, - xpTriple, xpWrap ) -- Local imports. import TSN.Codegen ( tsn_codegen_config, tsn_db_field_namer ) -- Used in a test +import TSN.Database ( insert_or_select ) import TSN.DbImport ( DbImport(..), ImportResult(..), run_dbmigrate ) +import TSN.Picklers ( xp_time_stamp ) +import TSN.Location ( Location(..), pickle_location ) import TSN.XmlImport ( XmlImport(..) ) -import Xml ( FromXml(..), pickle_unpickle, unpickleable ) +import Xml ( + FromXml(..), + ToDb(..), + pickle_unpickle, + unpickleable, + unsafe_read_document, + unsafe_unpickle ) - --- | The database type for teams as they show up in the news. +-- | The DTD to which this module corresponds. Used to invoke dbimport. -- -data NewsTeam = - NewsTeam { team_name :: String } - deriving (Eq, Show) +dtd :: String +dtd = "newsxml.dtd" --- | This is needed to define the XmlImport instance for NewsTeam; it --- basically says that the DB representation is the same as the XML --- representation. --- -instance FromXml NewsTeam where - type Db NewsTeam = NewsTeam - from_xml = id --- | Allow us to call 'insert_xml' on the XML representation of --- NewsTeams. -- -instance XmlImport NewsTeam - - --- | Mapping between News records and NewsTeam records in the --- database. We don't name the fields because we don't use the names --- explicitly; that means we have to give them nice database names --- via groundhog. --- -data News_NewsTeam = News_NewsTeam - (DefaultKey News) - (DefaultKey NewsTeam) - - --- | The database type for locations as they show up in the news. --- -data NewsLocation = - NewsLocation { - city :: Maybe String, - state :: Maybe String, - country :: String } - deriving (Eq, Show) - --- | This is needed to define the XmlImport instance for NewsLocation; it --- basically says that the DB representation is the same as the XML --- representation. --- -instance FromXml NewsLocation where - type Db NewsLocation = NewsLocation - from_xml = id - --- | Allow us to call 'insert_xml' on the XML representation of --- NewsLocations. --- -instance XmlImport NewsLocation - - --- | Mapping between News records and NewsLocation records in the --- database. We don't name the fields because we don't use the names --- explicitly; that means we have to give them nice database names --- via groundhog. +-- DB/XML Data types -- -data News_NewsLocation = News_NewsLocation - (DefaultKey News) - (DefaultKey NewsLocation) +-- * News/Message --- | The msg_id child of contains an event_id attribute; we --- embed it into the 'News' type. We (pointlessly) use the "db_" --- prefix here so that the two names don't collide on "id" when +-- | The msg_id child of \ contains an event_id attribute; we +-- embed it into the 'News' type. We (pointlessly) use the \"db_\" +-- prefix here so that the two names don't collide on \"id\" when -- Groundhog is creating its fields using our field namer. +-- data MsgId = MsgId { db_msg_id :: Int, @@ -134,7 +108,7 @@ data MsgId = deriving (Data, Eq, Show, Typeable) --- | The XML representation of a news item (message). +-- | The XML representation of a news item (\). -- data Message = Message { @@ -145,12 +119,12 @@ data Message = xml_sport :: String, xml_url :: Maybe String, xml_teams :: [NewsTeam], - xml_locations :: [NewsLocation], + xml_locations :: [Location], xml_sms :: String, xml_editor :: Maybe String, xml_text :: Maybe String, -- Text and continue seem to show up in pairs, xml_continue :: Maybe String, -- either both present or both missing. - xml_time_stamp :: String } + xml_time_stamp :: UTCTime } deriving (Eq, Show) @@ -160,68 +134,149 @@ data Message = -- data News = News { + db_xml_file_id :: Int, db_mid :: MsgId, db_sport :: String, db_url :: Maybe String, db_sms :: String, db_editor :: Maybe String, db_text :: Maybe String, - db_continue :: Maybe String } + db_continue :: Maybe String, + db_time_stamp :: UTCTime } deriving (Data, Eq, Show, Typeable) + +instance ToDb Message where + -- | The database representation of 'Message' is 'News'. + type Db Message = News + -- | Convert the XML representation 'Message' to the database -- representation 'News'. -- instance FromXml Message where - type Db Message = News - -- | We use a record wildcard so GHC doesn't complain that we never -- used the field names. -- - from_xml Message{..} = News { db_mid = xml_mid, + from_xml Message{..} = News { db_xml_file_id = xml_xml_file_id, + db_mid = xml_mid, db_sport = xml_sport, db_url = xml_url, db_sms = xml_sms, db_editor = xml_editor, db_text = xml_text, - db_continue = xml_continue } + db_continue = xml_continue, + db_time_stamp = xml_time_stamp } --- | This lets us call 'insert_xml' on a 'Message'. +-- | This lets us insert the XML representation 'Message' directly. -- instance XmlImport Message + +-- * NewsTeam + +-- | The database/XML type for teams as they show up in the news. We +-- can't reuse the representation from "TSN.Team" because they +-- require a team id. We wouldn't want to make the team ID optional +-- and then insert a team with no id, only to find the same team +-- later with an id and be unable to update the record. (We could +-- add the update logic, but it would be more trouble than it's +-- worth.) +-- +data NewsTeam = + NewsTeam { team_name :: String } + deriving (Eq, Show) + + + +-- * News_NewsTeam + +-- | Mapping between News records and NewsTeam records in the +-- database. We don't name the fields because we don't use the names +-- explicitly; that means we have to give them nice database names +-- via groundhog. +-- +data News_NewsTeam = News_NewsTeam + (DefaultKey News) + (DefaultKey NewsTeam) + + +-- * News_Location + +-- | Mapping between 'News' records and 'Location' records in the +-- database. We don't name the fields because we don't use the names +-- explicitly; that means we have to give them nice database names +-- via groundhog. +-- +data News_Location = News_Location + (DefaultKey News) + (DefaultKey Location) + + + + +-- | Some newsxml documents contain two \ elements in a row, +-- violating the DTD. The second one has always been empty, but it's +-- irrelevant: we can't parse these, and would like to detect them +-- in order to report the fact that the busted document is +-- unsupported. +-- +-- This function detects whether two \ elements appear in a +-- row, as siblings. +-- +has_only_single_sms :: XmlTree -> Bool +has_only_single_sms xmltree = + case elements of + [] -> True + _ -> False + where + parse :: XmlTree -> [XmlTree] + parse = runLA $ hasName "/" + /> hasName "message" + >>> addNav + >>> descendantAxis + >>> filterAxis (hasName "SMS") + >>> followingSiblingAxis + >>> remNav + >>> hasName "SMS" + + elements = parse xmltree + + +-- +-- * Database code +-- + -- | Define 'dbmigrate' and 'dbimport' for 'Message's. The import is -- slightly non-generic because of our 'News_NewsTeam' and --- 'News_NewsLocation' join tables. +-- 'News_Location' join tables. -- instance DbImport Message where dbmigrate _ = run_dbmigrate $ do - migrate (undefined :: NewsTeam) - migrate (undefined :: NewsLocation) + migrate (undefined :: Location) migrate (undefined :: News) + migrate (undefined :: NewsTeam) migrate (undefined :: News_NewsTeam) - migrate (undefined :: News_NewsLocation) + migrate (undefined :: News_Location) dbimport message = do -- Insert the message and acquire its primary key (unique ID) news_id <- insert_xml message - -- And insert each one into its own table. We use insert_xml_or_select - -- because we know that most teams will already exist, and we - -- want to get back the id for the existing team when - -- there's a collision. - nt_ids <- mapM insert_xml_or_select (xml_teams message) + -- Now insert the teams. We use insert_or_select because we know + -- that most teams will already exist, and we want to get back the + -- id for the existing team when there's a collision. + nt_ids <- mapM insert_or_select (xml_teams message) -- Now that the teams have been inserted, create -- news__news_team records mapping beween the two. let news_news_teams = map (News_NewsTeam news_id) nt_ids mapM_ insert_ news_news_teams - -- Do all of that over again for the NewsLocations. - loc_ids <- mapM insert_xml_or_select (xml_locations message) - let news_news_locations = map (News_NewsLocation news_id) loc_ids + -- Do all of that over again for the Locations. + loc_ids <- mapM insert_or_select (xml_locations message) + let news_news_locations = map (News_Location news_id) loc_ids mapM_ insert_ news_news_locations return ImportSucceeded @@ -240,15 +295,6 @@ mkPersist defaultCodegenConfig [groundhog| type: constraint fields: [team_name] -- entity: NewsLocation - dbName: news_locations - constructors: - - name: NewsLocation - uniques: - - name: unique_news_location - type: constraint - fields: [city, state, country] - |] @@ -256,14 +302,19 @@ mkPersist defaultCodegenConfig [groundhog| -- use our own codegen to peel those off before naming the columns. mkPersist tsn_codegen_config [groundhog| - entity: News - dbName: news constructors: - name: News + uniques: + - name: unique_news + type: constraint + # Prevent multiple imports of the same message. + fields: [db_xml_file_id] fields: - name: db_mid embeddedType: - {name: msg_id, dbName: msg_id} - {name: event_id, dbName: event_id} + - embedded: MsgId fields: - name: db_msg_id @@ -271,7 +322,6 @@ mkPersist tsn_codegen_config [groundhog| - name: db_event_id dbName: event_id - - entity: News_NewsTeam dbName: news__news_teams constructors: @@ -279,21 +329,33 @@ mkPersist tsn_codegen_config [groundhog| fields: - name: news_NewsTeam0 # Default created by mkNormalFieldName dbName: news_id + reference: + onDelete: cascade - name: news_NewsTeam1 # Default created by mkNormalFieldName dbName: news_teams_id + reference: + onDelete: cascade -- entity: News_NewsLocation - dbName: news__news_locations +- entity: News_Location + dbName: news__locations constructors: - - name: News_NewsLocation + - name: News_Location fields: - - name: news_NewsLocation0 # Default created by mkNormalFieldName + - name: news_Location0 # Default created by mkNormalFieldName dbName: news_id - - name: news_NewsLocation1 # Default created by mkNormalFieldName - dbName: news_locations_id + reference: + onDelete: cascade + - name: news_Location1 # Default created by mkNormalFieldName + dbName: locations_id + reference: + onDelete: cascade |] +-- +-- XML Picklers +-- + -- | Convert a 'NewsTeam' to/from XML. -- pickle_news_team :: PU NewsTeam @@ -320,20 +382,6 @@ pickle_msg_id = to_tuple m = (db_msg_id m, db_event_id m) --- | Convert a 'NewsLocation' to/from XML. --- -pickle_location :: PU NewsLocation -pickle_location = - xpElem "location" $ - xpWrap (from_tuple, to_tuple) $ - xpTriple (xpOption (xpElem "city" xpText)) - (xpOption (xpElem "state" xpText)) - (xpElem "country" xpText) - where - from_tuple = - uncurryN NewsLocation - to_tuple l = (city l, state l, country l) - -- | Convert a 'Message' to/from XML. -- @@ -353,7 +401,7 @@ pickle_message = (xpOption (xpElem "Editor" xpText)) (xpOption (xpElem "text" xpText)) pickle_continue - (xpElem "time_stamp" xpText) + (xpElem "time_stamp" xp_time_stamp) where from_tuple = uncurryN Message to_tuple m = (xml_xml_file_id m, -- Verbose, @@ -398,8 +446,10 @@ news_tests = testGroup "News tests" [ test_news_fields_have_correct_names, + test_on_delete_cascade, test_pickle_of_unpickle_is_identity, - test_unpickle_succeeds ] + test_unpickle_succeeds, + test_sms_detected_correctly ] -- | Make sure our codegen is producing the correct database names. @@ -420,13 +470,20 @@ test_news_fields_have_correct_names = map (\x -> tsn_db_field_namer "herp" "derp" 8675309 x 90210) field_names actual :: [String] - actual = ["mid", "sport", "url", "sms", "editor", "text", "continue"] + actual = ["xml_file_id", + "mid", + "sport", + "url", + "sms", + "editor", + "text", + "continue"] check (x,y) = (x @?= y) -- | If we unpickle something and then pickle it, we should wind up --- with the same thing we started with. WARNING: succeess of this +-- with the same thing we started with. WARNING: success of this -- test does not mean that unpickling succeeded. -- test_pickle_of_unpickle_is_identity :: TestTree @@ -456,3 +513,57 @@ test_unpickle_succeeds = testGroup "unpickle tests" actual <- unpickleable path pickle_message let expected = True actual @?= expected + + +-- | Make sure everything gets deleted when we delete the top-level +-- record. +-- +test_on_delete_cascade :: TestTree +test_on_delete_cascade = testGroup "cascading delete tests" + [ check "deleting news deletes its children" + "test/xml/newsxml.xml" + 4 -- 2 news_teams and 2 news_locations that should remain. + ] + where + check desc path expected = testCase desc $ do + news <- unsafe_unpickle path pickle_message + let a = undefined :: Location + let b = undefined :: News + let c = undefined :: NewsTeam + let d = undefined :: News_NewsTeam + let e = undefined :: News_Location + actual <- withSqliteConn ":memory:" $ runDbConn $ do + runMigration silentMigrationLogger $ do + migrate a + migrate b + migrate c + migrate d + migrate e + _ <- dbimport news + deleteAll b + count_a <- countAll a + count_b <- countAll b + count_c <- countAll c + count_d <- countAll d + count_e <- countAll e + return $ count_a + count_b + count_c + count_d + count_e + actual @?= expected + + +-- | We want to make sure the single-SMS documents and the multi-SMS +-- documents are identified correctly. +-- +test_sms_detected_correctly :: TestTree +test_sms_detected_correctly = + testGroup "newsxml SMS count determined correctly" + [ check "test/xml/newsxml.xml" + "single SMS detected correctly" + True, + check "test/xml/newsxml-multiple-sms.xml" + "multiple SMS detected correctly" + False ] + where + check path desc expected = testCase desc $ do + xmltree <- unsafe_read_document path + let actual = has_only_single_sms xmltree + actual @?= expected