Use /tmp for running dbmate

Reviewed-on: #2

.gitea/workflows/build.yml

@@ -0,0 +1,27 @@
+name: build
+on:
+  pull_request:
+jobs:
+  build:
+    container:
+      image: node:24-bookworm
+      env:
+        USER: node
+        HOME: /home/node
+        NIX_CONFIG: experimental-features = nix-command flakes
+      volumes:
+        - /var/lib/pbri/docker/nix_runner_nix:/nix
+        - /var/lib/pbri/docker/nix_runner_etc:/etc/nix
+        - /var/lib/pbri/docker/nix_runner_home_node:/home/node
+      options: --user 1000:1000
+    steps:
+      - name: Prep nix env
+        run: |
+          if [ ! -f ~/.nix-profile/etc/profile.d/nix.sh ]; then
+            curl -L https://nixos.org/nix/install | bash -s -- --no-daemon
+          fi
+          . ~/.nix-profile/etc/profile.d/nix.sh 
+          env >> "$GITHUB_ENV"
+      - name: Checkout repo
+        uses: actions/checkout@v4
+      - run: nix build --log-format raw --print-build-logs .#yore

.gitea/workflows/deliver.yml

@@ -0,0 +1,40 @@
+name: deliver
+on:
+  push:
+    tags:
+      - 'v*'
+jobs:
+  deliver:
+    container:
+      image: node:24-bookworm
+      env:
+        USER: node
+        HOME: /home/node
+        NIX_CONFIG: experimental-features = nix-command flakes
+      volumes:
+        - /var/lib/pbri/docker/nix_runner_nix:/nix
+        - /var/lib/pbri/docker/nix_runner_etc:/etc/nix
+        - /var/lib/pbri/docker/nix_runner_home_node:/home/node
+      options: --user 1000:1000
+    steps:
+      - name: Prep nix env
+        run: |
+          if [ ! -f ~/.nix-profile/etc/profile.d/nix.sh ]; then
+            curl -L https://nixos.org/nix/install | bash -s -- --no-daemon
+          fi
+          . ~/.nix-profile/etc/profile.d/nix.sh
+          env >> "$GITHUB_ENV"
+      - name: Checkout repo
+        uses: actions/checkout@v4
+      - name: Get image meta
+        run: nix build .#yore-meta -o /tmp/yore-meta
+      - name: Version check
+        run: |
+          VERSION=$(cat /tmp/yore-meta/version)
+          [ "$GITHUB_REF_NAME" = v"$VERSION" ]
+      - run: nix build --log-format raw --print-build-logs .#image -o image.tar.gz
+      - run: gunzip -c image.tar.gz > image.tar
+      - run: nix run .#crane -- auth login git.pbrinkmeier.de -u paul -p "$PASSWORD"
+        env:
+          PASSWORD: "${{ secrets.REGISTRY_PASSWORD }}"
+      - run: nix run .#crane -- push image.tar $(cat /tmp/yore-meta/image-tag)

README.md

@@ -12,7 +12,7 @@ The rest of this section assumes you're in the dev shell.
 
 ### Running Postgres
 
-`scripts/with-db $COMMAND` will create a Postgres cluster in a temporary directory with the credentials `user=yore-test dbname=yore-test port=5433`, start the Postgres server, run `$COMMAND` and then stop the Postgres erver.
+`scripts/with-db $COMMAND` will create a Postgres cluster in a temporary directory with the credentials `user=yore-test dbname=yore-test port=5433`, start the Postgres server, apply migrations using `dbmate`, run `$COMMAND` and then stop the Postgres server.
 On Linux, you need to have write permissions vor `/var/run/postgresql` to run it.
 You can use this to run the tests in an entirely fresh database, e.g.:
 
@@ -24,6 +24,8 @@ You can also use it to run a database for development by choosing a `$COMMAND` t
 
 ```
 BASE_DIR=./pgdata scripts/with-db psql "dbname=yore-test port=5433 user=yore-test"
+# cabal repl with env vars set
+BASE_DIR=./pgdata scripts/with-db cabal run
 ```
 
 By setting `$BASE_DIR` you can persist the database for later runs.
@@ -39,6 +41,10 @@ By setting `$BASE_DIR` you can persist the database for later runs.
 
 ## TODO
 
-- CI
-- Docker container (in flake)
-- try fourmolu
+- [x] CI
+- [x] Docker container (in flake)
+- [x] try fourmolu
+- [x] Test leap second handling in the `time` package
+- [ ] Fix container shutdown delay
+- [ ] Fix `responseFile` always returning 200
+- [ ] use queue to sync logging of indexer and main thread

app/Main.hs

@@ -1,37 +1,42 @@
 {-# LANGUAGE DataKinds #-}
+{-# LANGUAGE ExplicitNamespaces #-}
 {-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE OverloadedRecordDot #-}
 {-# LANGUAGE OverloadedStrings #-}
 {-# LANGUAGE PatternSynonyms #-}
+{-# LANGUAGE QuantifiedConstraints #-}
+{-# LANGUAGE StandaloneDeriving #-}
+{-# LANGUAGE TypeSynonymInstances #-}
 
 module Main (main) where
 
+import Control.Concurrent (forkIO)
+import Control.Monad (forM, forM_)
 import Control.Monad.IO.Class (liftIO)
-import Control.Monad.Trans.Except (ExceptT (..), runExceptT)
+import Control.Monad.Trans.Except (ExceptT (..), catchE, runExceptT)
+import Data.Bifunctor (Bifunctor (..))
 import Data.Proxy (Proxy (..))
-import Data.Time (Day, toGregorian)
+import Data.Text (Text)
+import Data.Time (Day, LocalTime (..), TimeOfDay (..), ZonedTime (..), addDays, nominalDiffTimeToSeconds, toGregorian)
 import GHC.Generics (Generic)
-import Network.HTTP.Media ((//), (/:))
-import Network.HTTP.Types (Status (..), mkStatus, status200)
+import Network.HTTP.Types (Status (..))
+import Network.Wai (Middleware, Request (..), responseStatus)
 import Servant
-  ( Accept (..)
-  , Capture
+  ( Capture
   , Get
   , Handler
   , MimeRender (..)
-  , Raw
   , ServerError (..)
   , ServerT
-  , Tagged (..)
   , err404
   , err500
   , hoistServer
-  , runHandler
   , serve
   , (:<|>) (..)
   , (:>)
   , pattern MkHandler
   )
+import System.Exit (exitFailure)
 import System.FilePath ((</>))
 import Text.Blaze.Html.Renderer.Utf8 (renderHtml)
 import Text.Blaze.Html5 ((!))
@@ -42,47 +47,102 @@ import qualified Data.ByteString.Lazy as LBS
 import qualified Data.Text as Text
 import qualified Data.Text.Encoding as Encoding
 import qualified Database.PostgreSQL.Opium as Opium
-import Network.Wai (Application, Request (..), Response, responseFile, responseLBS)
 import qualified Network.Wai.Handler.Warp as Warp
 import qualified Text.Blaze.Html5 as H
 import qualified Text.Blaze.Html5.Attributes as A
 
-import Yore.DB (DayFile (..))
+import Envy (type (=@!), type (=@@), type (?))
+import Yore.DB (DB, DayFile (..))
+import Yore.Download (downloadInto)
 import Yore.Error (Error (..))
+import Yore.Schedule (schedule)
+import Yore.Scrape (Issue (..), getIssuesByDay)
+import Yore.Servant (GetSendfile, HTML, Sendfile (..))
 import Yore.Time (addYears, getCurrentDay, getSecondsUntilMidnight)
 
+import Data.Fixed (Pico)
+import Data.Time.Clock.POSIX (getPOSIXTime)
 import qualified Envy
 import qualified Yore.DB as DB
 import qualified Yore.Log as Log
+import System.IO (hSetBuffering, BufferMode (..), stdout)
 
-data Config f = Config
-  { port :: Envy.Env f (Envy.Optional Integer 3000)
-  , downloadDir :: Envy.Env f (Envy.Optional String "download")
-  , factor :: Envy.Env f (Envy.Required Rational)
-  , factorFallback :: Envy.Env f (Envy.Optional Rational '(1, 1))
+data ConfigT f = Config
+  { yorePort :: f =@@ Int ? 3000
+  , yoreDownloadDir :: f =@@ FilePath ? "./download"
+  , yoreDb :: f =@! Text
   }
   deriving (Generic)
 
+type Config = ConfigT Envy.Value
+
+deriving instance Show Config
+
 main :: IO ()
 main = do
-  print . port =<< Envy.load @Config
-  print . downloadDir =<< Envy.load @Config
-  print . factor =<< Envy.load @Config
-  let
-    settings =
-      Warp.setLogger logger $
-        Warp.setPort 3000 $
-          Warp.setOnException onException Warp.defaultSettings
+  hSetBuffering stdout NoBuffering
 
-  db <- DB.initDB "host=localhost user=yore-test port=5433 dbname=yore-test"
-  Warp.runSettings settings $ serve (Proxy :: Proxy API) $ hoistServer (Proxy :: Proxy API) nt $ server db
+  cfg <-
+    Envy.load @ConfigT >>= \case
+      Left errs -> do
+        forM_ errs $ Log.error . printf "failed to read config: %s"
+        exitFailure
+      Right c ->
+        pure c
+
+  db <- DB.initDB cfg.yoreDb
+
+  _ <- runExceptT $ doIndex cfg db
+
+  _ <- forkIO $ runIndexer cfg db
+  runServer cfg db
+
+runIndexer :: Config -> DB -> IO ()
+runIndexer cfg db =
+  schedule shouldRunAt $ runExceptT $ doIndex cfg db
   where
-    logger req status _ = do
-      Log.info $
-        printf
-          "%d %s"
-          (statusCode status)
-          (requestLine req)
+    shouldRunAt zt = case localTimeOfDay $ zonedTimeToLocalTime zt of
+      TimeOfDay 3 0 _ -> True
+      _ -> False
+
+doIndex :: Config -> DB -> ExceptT Error IO ()
+doIndex cfg db =
+  catchE
+    ( DB.withTransaction db $ \conn -> do
+        forM_ [0, 1, 2] $ \i -> do
+          dayThen <- getTodayWithOffset (-100) i
+          indexDay cfg dayThen conn
+    )
+    (Log.error . show)
+
+runServer :: Config -> DB -> IO ()
+runServer cfg db =
+  Warp.runSettings settings $ loggerMiddleware $ serve (Proxy @API) $ hoistServer (Proxy @API) nt $ server cfg db
+  where
+    settings =
+      foldr
+        ($)
+        Warp.defaultSettings
+        [ Warp.setPort cfg.yorePort
+        , Warp.setBeforeMainLoop $ Log.info $ printf "listening on port %d" cfg.yorePort
+        , Warp.setOnException onException
+        ]
+
+    logger :: Request -> Status -> Pico -> IO ()
+    logger req status s = do
+      Log.info $ printf "%d %s %ss" (statusCode status) (requestLine req) (formatMetric s)
+
+    loggerMiddleware :: Middleware
+    loggerMiddleware app req respond = do
+      begin <- getPOSIXTime
+      app req $ \res -> do
+        rr <- respond res
+        end <- getPOSIXTime
+        logger req (responseStatus res) (nominalDiffTimeToSeconds (end - begin))
+        pure rr
+
+    onException mbReq ex = do
+      Log.error $ printf "unhandled exception%s: %s%s" (maybe "" ((" in " ++) . requestLine) mbReq) (show ex)
 
     requestLine :: Request -> String
     requestLine req =
@@ -91,8 +151,17 @@ main = do
         (BS8.unpack $ requestMethod req)
         (BS8.unpack $ rawPathInfo req <> rawQueryString req)
 
-    onException mbReq ex = do
-      Log.error $ printf "unhandled exception%s: %s%s" (maybe "" ((" in " ++) . requestLine) mbReq) (show ex)
+formatMetric :: Pico -> String
+formatMetric x = go (-9) prefixes
+  where
+    prefixes = ["p", "n", "u", "m", "", "k", "M", "G"]
+
+    go :: Int -> [String] -> String
+    go _ [] = show x
+    go e (p : ps)
+      | x < 10 ^^ e =
+          show @Int (floor $ x * 10 ^^ (-e + 3)) ++ p
+      | otherwise = go (e + 3) ps
 
 nt :: ExceptT Error IO a -> Handler a
 nt action = MkHandler $ do
@@ -104,11 +173,6 @@ nt action = MkHandler $ do
     Right x ->
       pure $ Right x
 
-type API =
-  Get '[HTML] RootModel
-    :<|> "today" :> "issue" :> Capture "issue" Int :> Get '[HTML] RootModel
-    :<|> "api" :> "today" :> "issue" :> Capture "issue" Int :> "fz.pdf" :> Raw
-
 toServerError :: Error -> ServerError
 toServerError = \case
   ConnectionError msg ->
@@ -123,39 +187,36 @@ toServerError = \case
 encodeUtf8LBS :: String -> LBS.ByteString
 encodeUtf8LBS = LBS.fromStrict . Encoding.encodeUtf8 . Text.pack
 
-handlerToRaw :: ExceptT Error IO Response -> Tagged t Application
-handlerToRaw handler = Tagged $ \_ respond -> do
-  r <- runHandler $ nt handler
-  case r of
-    Left e ->
-      respond $ responseLBS (mkStatus e.errHTTPCode $ BS8.pack e.errReasonPhrase) e.errHeaders e.errBody
-    Right response ->
-      respond response
+type API =
+  Get '[HTML] RootModel
+    :<|> "today" :> "issue" :> Capture "issue" Int :> Get '[HTML] RootModel
+    :<|> "api" :> "today" :> "issue" :> Capture "issue" Int :> "fz.pdf" :> GetSendfile
 
-server :: DB.DB -> ServerT API (ExceptT Error IO)
-server db = rootR :<|> todayR :<|> apiTodayR
+server :: Config -> DB -> ServerT API (ExceptT Error IO)
+server cfg db = rootR :<|> todayR :<|> apiTodayR
   where
     rootR = todayR 0
 
     todayR issue = do
-      dateThen <- ExceptT get100YearsAgo
-      count <- ExceptT $ DB.withConn db $ DB.getNumberOfIssues dateThen
-      dayFile <- ExceptT $ DB.withConn db $ DB.getDayFileByIssue dateThen issue
+      dateThen <- getTodayWithOffset (-100) 0
+      count <- DB.withConn db $ DB.getNumberOfIssues dateThen
+      dayFile <- DB.withConn db $ DB.getDayFileByIssue dateThen issue
       pure $ RootModel dateThen dayFile issue count
 
-    apiTodayR issue = handlerToRaw $ do
-      dateThen <- ExceptT get100YearsAgo
-      dayFile <- ExceptT $ DB.withConn db $ DB.getDayFileByIssue dateThen issue
-      let fullPath = "download" </> dayFile.relative_path
+    apiTodayR issue = do
+      dateThen <- getTodayWithOffset (-100) 0
+      dayFile <- DB.withConn db $ DB.getDayFileByIssue dateThen issue
+      let fullPath = cfg.yoreDownloadDir </> dayFile.relative_path
       secondsUntilMidnight <- liftIO getSecondsUntilMidnight
-      pure $
-        responseFile
-          status200
-          [ ("content-type", "application/pdf")
-          , ("cache-control", BS8.pack $ printf "public, max-age=%d" secondsUntilMidnight)
-          ]
-          fullPath
-          Nothing
+      pure
+        Sendfile
+          { headers =
+              [ ("content-type", "application/pdf")
+              , ("cache-control", BS8.pack $ printf "public, max-age=%d" secondsUntilMidnight)
+              ]
+          , path =
+              fullPath
+          }
 
 data RootModel = RootModel Day DB.DayFile Int Int
 
@@ -191,11 +252,31 @@ instance MimeRender HTML RootModel where
 
 -- Utils
 
-data HTML
+getTodayWithOffset :: Integer -> Integer -> ExceptT Error IO Day
+getTodayWithOffset yearOffset dayOffset =
+  ExceptT $
+    first (GenericError . (Text.pack (printf "can't go back %d years and go forward %d days: " (-yearOffset) dayOffset) <>))
+      . addYears yearOffset
+      . addDays dayOffset
+      <$> getCurrentDay
 
-instance Accept HTML where
-  contentType _ = "text" // "html" /: ("charset", "utf-8")
-
-get100YearsAgo :: IO (Either Error Day)
-get100YearsAgo =
-  either (Left . GenericError . ("can't go back 100 years: " <>)) Right . addYears (-100) <$> getCurrentDay
+indexDay :: Config -> Day -> Opium.Connection -> ExceptT Error IO ()
+indexDay cfg dayThen conn = do
+  -- Transaction-level lock released automatically after transaction
+  DB.lift $ Opium.execute_ "SELECT pg_advisory_xact_lock(42);" conn
+  mbDi <- DB.readDayIndex dayThen conn
+  case mbDi of
+    Just _ ->
+      Log.info $ printf "index for %s already exists." (show dayThen)
+    Nothing -> do
+      Log.info $ printf "scraping issues for %s" (show dayThen)
+      issues <- liftIO $ getIssuesByDay dayThen
+      paths <- forM issues $ \issue -> do
+        Log.info $ printf "downloading %s" issue.url
+        path <- downloadInto cfg.yoreDownloadDir issue.url
+        pure (issue.label, path)
+      Log.info "creating DB entries"
+      dayIndex <- DB.createDayIndex dayThen conn
+      forM_ paths $ \(text, url) ->
+        DB.createDayFile dayIndex.day_index_id text url conn
+      Log.info "done."

db/schema.sql

@@ -1,3 +1,8 @@
+\restrict CAC30PXzQJosjH02Spia1vzj9Ui3LB1Cq2HI8r0A4eu33HP8ngxbbM2SzRfTj0c
+
+-- Dumped from database version 17.6
+-- Dumped by pg_dump version 17.6
+
 SET statement_timeout = 0;
 SET lock_timeout = 0;
 SET idle_in_transaction_session_timeout = 0;
@@ -124,6 +129,8 @@ ALTER TABLE ONLY yore.day_file
 -- PostgreSQL database dump complete
 --
 
+\unrestrict CAC30PXzQJosjH02Spia1vzj9Ui3LB1Cq2HI8r0A4eu33HP8ngxbbM2SzRfTj0c
+
 
 --
 -- Dbmate schema migrations

flake.lock

@@ -40,11 +40,11 @@
     },
     "nixpkgs": {
       "locked": {
-        "lastModified": 1752536923,
-        "narHash": "sha256-fdgPZR7VFSSRIQKOJLcs3qCJBWM64Uak0gAGtWTYAd8=",
+        "lastModified": 1758518877,
+        "narHash": "sha256-1K1c6vNr/N1PfM77DqKINXQ/oCxA0W1ZJ80sfNzWu0w=",
         "owner": "nixos",
         "repo": "nixpkgs",
-        "rev": "c665e4d918eda5d78a175ed8d300809c44932160",
+        "rev": "033cc13ab4c1a60b3527b242755aa625001ff1f9",
         "type": "github"
       },
       "original": {
@@ -64,11 +64,11 @@
         ]
       },
       "locked": {
-        "lastModified": 1756064767,
-        "narHash": "sha256-SaJK9VYc/oF5aBSMSW+We8TggER4+uTAHvc2ztnLpYo=",
+        "lastModified": 1759790870,
+        "narHash": "sha256-rEsaO38v2wvAThPzgbYPgnv29mu6GxTYLjgci3GYaO4=",
         "ref": "main",
-        "rev": "af16429d827816cf9bec69026e13f50acc3d2691",
-        "revCount": 65,
+        "rev": "4db509c1b52ee7b5012abfef5e11e62beab27a16",
+        "revCount": 66,
         "type": "git",
         "url": "https://git.pbrinkmeier.de/paul/opium"
       },

flake.nix

@@ -26,11 +26,18 @@
         };
 
         # For building the package, we use only the files not ignored by Git as inputs.
-        # Also, flake.nix and flake.lock are not included to avoid rebuilds when
+        # Also, flake.nix, flake.lock and .gitea are not included to avoid rebuilds when
         # working on them.
         src = pkgs.lib.cleanSourceWith {
-          src = gitignore.lib.gitignoreSource ./.;
-          filter = path: type: builtins.elem (builtins.baseNameOf path) [ "flake.nix" "flake.lock" ];
+          src = ./.;
+          filter = gitignore.lib.gitignoreFilterWith {
+            basePath = ./.;
+            extraRules = ''
+              flake.nix
+              flake.lock
+              .gitea
+            '';
+          };
         };
 
         yore = pkgs.lib.pipe
@@ -51,16 +58,44 @@
             # Return only the bin folder to curb image size
             compose.justStaticExecutables
           ];
+
+        image =
+          pkgs.dockerTools.buildImage {
+            name = "git.pbrinkmeier.de/paul/yore";
+            tag = yore.version;
+            copyToRoot = [ pkgs.cacert ];
+            config = {
+              Cmd = [ "${entrypoint}/bin/run" ];
+              StopSignal = "SIGINT";
+            };
+          };
+
+        entrypoint = pkgs.writeShellApplication {
+          name = "run";
+          runtimeInputs = [ pkgs.coreutils pkgs.dbmate yore ];
+          text = ''
+            set -Eeuo pipefail
+
+            mkdir -p /tmp/yore-workspace/db
+            cp -r ${./db}/* /tmp/yore-workspace/db
+            cd /tmp/yore-workspace/db
+            dbmate up
+
+            # Replace bash and inherit PID 1
+            cd /
+            exec yore
+          '';
+        };
       in {
         packages = rec {
-          docker =
-            pkgs.dockerTools.buildImage {
-              name = "git.pbrinkmeier.de/paul/yore";
-              tag = yore.version;
-              config.Cmd = [ "${yore}/bin/yore" ];
-            };
-
           default = yore;
+          inherit yore image;
+          yore-meta = pkgs.runCommand "yore-meta" {} ''
+            mkdir -p $out
+            echo -n ${yore.version} > $out/version
+            echo -n ${image.drvAttrs.imageName}:${image.drvAttrs.imageTag} > $out/image-tag
+          '';
+
 
           opium_ = opium.packages.${system}.opium;
         };
@@ -82,5 +117,11 @@
               overrides = addOpium;
             }).env;
         };
+        apps = {
+          crane = {
+            type = "app";
+            program = "${pkgs.crane}/bin/crane";
+          };
+        };
       });
 }

scripts/with-db

@@ -5,19 +5,27 @@ if [ -z "$BASE_DIR" ]; then
 fi
 export PGDATA="$BASE_DIR/pgdata"
 
+export DB_USER=yore-test
+export DB_DBNAME=yore-test
+export DB_PORT=5433
+
 # On error or exit: Stop the DB server
 stop_db_server() {
-    pg_ctl stop -o "-p 5433" --wait
+    pg_ctl stop -o "-p ${DB_PORT}" --wait
 }
 trap stop_db_server EXIT
 
 if [ ! -s "$PGDATA/PG_VERSION" ]; then
-    initdb --username=yore-test --auth=trust || true
+    initdb --username="${DB_USER}" --auth=trust || true
     DO_CREATE_DB=1
 fi
 pg_ctl start -o "-p 5433" --wait
 if [ -n "$DO_CREATE_DB" ]; then
-    createdb -p 5433 yore-test --username=yore-test || true
+    createdb -p "${DB_PORT}" "${DB_DBNAME}" --username="${DB_USER}" || true
 fi
 
+DATABASE_URL="postgres://${DB_USER}@localhost:${DB_PORT}/${DB_DBNAME}?sslmode=disable" dbmate up
+
+export YORE_DB="host=localhost port=${DB_PORT} user=${DB_USER} dbname=${DB_DBNAME}" 
+
 "$@"

src/Envy.hs

@@ -1,53 +1,91 @@
 {-# LANGUAGE DataKinds #-}
 {-# LANGUAGE LambdaCase #-}
 {-# LANGUAGE TypeFamilies #-}
+{-# LANGUAGE TypeOperators #-}
+{-# LANGUAGE UndecidableInstances #-}
 
-module Envy (Env, Required, Optional, load) where
+module Envy
+  ( type (=@@)
+  , type (=@!)
+  , type (?)
+  , type (=@?)
+  , ReadEnvVar (..)
+  , Required
+  , Optional
+  , load
+  , ConfigVariant (..)
+  , MyNothing
+  , OptionalMaybe
+  ) where
 
 import Data.Bifunctor (first)
+import Data.Char (isLower, isUpper, toUpper)
 import Data.Kind (Type)
 import Data.Proxy (Proxy (..))
 import Data.Ratio ((%))
+import Data.Text (Text)
 import Data.Typeable (Typeable, typeRep)
 import GHC.Generics (C, D, Generic, K1 (..), M1 (..), Meta (..), Rec0, Rep, S, to, (:*:) (..))
-import GHC.TypeLits (KnownNat, KnownSymbol, natVal, symbolVal)
+import GHC.TypeLits (KnownNat, KnownSymbol, natVal, symbolVal, type (<=))
 import System.Environment (getEnvironment)
 import Text.Printf (printf)
 import Text.Read (readEither)
 
+import qualified Data.Text as Text
+
 data ConfigVariant = Spec | Value
 
 data Required t
 data Optional t d
 
-type family Env v t :: Type where
-  Env Spec t = t
-  Env Value (Required t) = t
-  Env Value (Optional t d) = t
+type family v =@@ t :: Type where
+  Spec =@@ t = t
+  Value =@@ (Required t) = t
+  Value =@@ (Optional t d) = t
+
+type OptionalMaybe t = Optional (Maybe t) MyNothing
+
+-- No associativity - these are not supposed to be chained
+infix 6 =@@
+infix 7 ?
+
+type v =@! t = v =@@ Required t
+type v =@? t = v =@@ OptionalMaybe t
+type t ? d = Optional t d
 
 load
-  :: forall (m :: ConfigVariant -> Type). (Generic (m Value), GFromEnv (Rep (m Spec)) (Rep (m Value))) => IO (m Value)
+  :: forall (m :: ConfigVariant -> Type)
+   . (Generic (m Value), GFromEnv (Rep (m Spec)) (Rep (m Value)))
+  => IO (Either [String] (m Value))
 load = do
   env <- getEnvironment
-  pure $ to $ gFromEnv @(Rep (m Spec)) @(Rep (m Value)) Proxy env
+  pure $ to <$> gFromEnv @(Rep (m Spec)) @(Rep (m Value)) Proxy env
 
 class GFromEnv spec value where
-  gFromEnv :: Proxy (spec c) -> [(String, String)] -> value c
+  gFromEnv :: Proxy (spec c) -> [(String, String)] -> Either [String] (value c)
 
 instance
   (GFromEnv i o)
   => GFromEnv (M1 D b i) (M1 D d o)
   where
-  gFromEnv Proxy env = M1 $ gFromEnv @i @o Proxy env
+  gFromEnv Proxy env = M1 <$> gFromEnv @i @o Proxy env
 
 instance
   (GFromEnv i o)
   => GFromEnv (M1 C b i) (M1 C d o)
   where
-  gFromEnv Proxy env = M1 $ gFromEnv @i @o Proxy env
+  gFromEnv Proxy env = M1 <$> gFromEnv @i @o Proxy env
 
 instance (GFromEnv i1 o1, GFromEnv i2 o2) => GFromEnv (i1 :*: i2) (o1 :*: o2) where
-  gFromEnv Proxy env = gFromEnv @i1 @o1 Proxy env :*: gFromEnv @i2 @o2 Proxy env
+  gFromEnv Proxy env =
+    case (eLhs, eRhs) of
+      (Left errs1, Right _) -> Left errs1
+      (Right _, Left errs2) -> Left errs2
+      (Left errs1, Left errs2) -> Left $ errs1 ++ errs2
+      (Right lhs, Right rhs) -> Right $ lhs :*: rhs
+    where
+      eLhs = gFromEnv @i1 @o1 Proxy env
+      eRhs = gFromEnv @i2 @o2 Proxy env
 
 instance
   (KnownSymbol sym, EnvVarSpec s t)
@@ -55,18 +93,17 @@ instance
       (M1 S (MetaSel (Just sym) su ss lz) (Rec0 s))
       (M1 S meta2 (Rec0 t))
   where
-  gFromEnv Proxy env = case decodeEnvVar @s @t Proxy varName $ lookup varName env of
-    Left err -> error err
-    Right value -> M1 $ K1 value
+  gFromEnv Proxy env =
+    M1 . K1 <$> first (: []) (decodeEnvVar @s @t Proxy varName (lookup varName env))
     where
-      varName = symbolVal $ Proxy @sym
+      varName = selectorNameToEnvVarName $ symbolVal $ Proxy @sym
 
 class EnvVarSpec specTy valueTy where
   decodeEnvVar :: Proxy specTy -> String -> Maybe String -> Either String valueTy
 
 instance (Typeable t, ReadEnvVar t) => EnvVarSpec (Required t) t where
   decodeEnvVar Proxy varName = \case
-    Nothing -> Left $ printf "%s is missing" varName
+    Nothing -> Left $ printf "env var %s is missing" varName
     Just str -> readEnvVarWithContext varName str
 
 instance (Typeable t, ReadEnvVar t, ProvidesDefault t d) => EnvVarSpec (Optional t d) t where
@@ -81,7 +118,7 @@ readEnvVarWithContext :: forall t. (Typeable t, ReadEnvVar t) => String -> Strin
 readEnvVarWithContext varName str =
   first (context ++) $ readEnvVar str
   where
-    context = printf "%s='%s' is not a valid %s: " varName str (show $ typeRep $ Proxy @t)
+    context = printf "env var %s='%s' is not a valid %s: " varName str (show $ typeRep $ Proxy @t)
 
 class ProvidesDefault t d where
   getDefault :: Proxy d -> t
@@ -89,17 +126,60 @@ class ProvidesDefault t d where
 instance ReadEnvVar Integer where
   readEnvVar = readEither
 
+-- TODO: Support negative integers (e.g. (Negate n))
 instance (KnownNat n) => ProvidesDefault Integer n where
   getDefault Proxy = natVal (Proxy @n)
 
+instance ReadEnvVar Int where
+  readEnvVar = readEither
+
+-- TODO: Support negative integers (e.g. (Negate n))
+-- This instance assumes that @Int@ is a 64-bit integer and enforces its range on the type level.
+instance (KnownNat n, n <= 9223372036854775807) => ProvidesDefault Int n where
+  getDefault Proxy = fromInteger $ natVal (Proxy @n)
+
 instance ReadEnvVar String where
   readEnvVar = Right
 
 instance (KnownSymbol s) => ProvidesDefault String s where
   getDefault Proxy = symbolVal (Proxy @s)
 
+instance ReadEnvVar Text where
+  readEnvVar = fmap Text.pack . readEnvVar
+
+instance (KnownSymbol s) => ProvidesDefault Text s where
+  getDefault Proxy = Text.pack $ getDefault (Proxy @s)
+
 instance ReadEnvVar Rational where
   readEnvVar = readEither
 
 instance (KnownNat num, KnownNat denom) => ProvidesDefault Rational '(num, denom) where
   getDefault Proxy = natVal (Proxy @num) % natVal (Proxy @denom)
+
+instance ReadEnvVar Double where
+  readEnvVar = readEither
+
+instance (KnownNat num, KnownNat denom) => ProvidesDefault Double '(num, denom) where
+  getDefault Proxy = fromRational $ getDefault @Rational @'(num, denom) Proxy
+
+instance (ReadEnvVar t) => ReadEnvVar (Maybe t) where
+  readEnvVar = fmap Just . readEnvVar
+
+instance (ProvidesDefault t d) => ProvidesDefault (Maybe t) (Just d) where
+  getDefault Proxy = Just $ getDefault @t @d Proxy
+
+data MyNothing
+
+instance ProvidesDefault (Maybe t) MyNothing where
+  getDefault Proxy = Nothing
+
+-- Converts camelCase names to UPPER_SNAKE_CASE.
+selectorNameToEnvVarName :: String -> String
+selectorNameToEnvVarName = go
+  where
+    go [] =
+      []
+    go (c0 : c1 : rest)
+      | isUpper c0 && isLower c1 = '_' : c0 : toUpper c1 : go rest
+    go (c0 : rest) =
+      toUpper c0 : go rest

src/Yore/DB.hs

@@ -9,6 +9,7 @@ module Yore.DB
   , Error (..)
   , initDB
   , withConn
+  , withTransaction
   , DayIndex (..)
   , DayFile (..)
   , createDayFile
@@ -17,9 +18,11 @@ module Yore.DB
   , readDayPaths
   , getDayFileByIssue
   , getNumberOfIssues
+  , lift
   ) where
 
 import Control.Concurrent (getNumCapabilities)
+import Control.Monad.Trans.Except (ExceptT (..), catchE, runExceptT, throwE)
 import Data.Functor.Identity (Identity (..))
 import Data.Pool (Pool, defaultPoolConfig, newPool, withResource)
 import Data.Text (Text)
@@ -28,6 +31,7 @@ import GHC.Generics (Generic)
 
 import qualified Database.PostgreSQL.Opium as Opium
 
+import Data.Bifunctor (Bifunctor (..))
 import Yore.Error (Error (..))
 
 newtype DB = DB (Pool (Either Opium.ConnectionError Opium.Connection))
@@ -45,13 +49,35 @@ initDB connString = do
           maxResources
       )
 
-withConn :: DB -> (Opium.Connection -> IO (Either Opium.Error a)) -> IO (Either Error a)
+-- TODO: This should probably also do something like bracket...
+-- For now let's assume no exceptions are thrown in f.
+withConn :: DB -> (Opium.Connection -> ExceptT Error IO a) -> ExceptT Error IO a
 withConn (DB connPool) f =
-  withResource connPool $ \case
+  ExceptT $ withResource connPool $ \case
     Left connectionError ->
       pure $ Left $ ConnectionError connectionError
     Right conn ->
-      either (Left . DBError) Right <$> f conn
+      runExceptT $ f conn
+
+withTransaction :: DB -> (Opium.Connection -> ExceptT Error IO a) -> ExceptT Error IO a
+withTransaction db f =
+  withConn db $ \conn ->
+    ( do
+        lift $ Opium.execute_ "BEGIN" conn
+        x <- f conn
+        lift $ Opium.execute_ "COMMIT" conn
+        pure x
+    )
+      `catchE` ( \err -> do
+                  -- rollback open transactions and release transaction level locks...
+                  lift $ Opium.execute_ "ROLLBACK" conn
+                  -- ...then rethrow the error so the app can handle it too.
+                  throwE err
+               )
+
+lift :: IO (Either Opium.Error a) -> ExceptT Error IO a
+lift f =
+  ExceptT $ first DBError <$> f
 
 data DayIndex = DayIndex
   { day_index_id :: Int
@@ -72,44 +98,47 @@ newtype CountResult = CountResult
   }
   deriving (Show, Generic, Opium.FromRow)
 
-createDayIndex :: Day -> Opium.Connection -> IO (Either Opium.Error DayIndex)
+createDayIndex :: Day -> Opium.Connection -> ExceptT Error IO DayIndex
 createDayIndex date =
   ex runIdentity
     . Opium.fetch
       "INSERT INTO yore.day_index (day) VALUES ($1) RETURNING *"
       (Identity date)
 
-createDayFile :: Int -> Text -> FilePath -> Opium.Connection -> IO (Either Opium.Error ())
+createDayFile :: Int -> Text -> FilePath -> Opium.Connection -> ExceptT Error IO ()
 createDayFile dayId label path =
-  Opium.execute
-    "INSERT INTO yore.day_file (day_index_id, label, relative_path) VALUES ($1, $2, $3)"
-    (dayId, label, path)
+  lift
+    . Opium.execute
+      "INSERT INTO yore.day_file (day_index_id, label, relative_path) VALUES ($1, $2, $3)"
+      (dayId, label, path)
 
-readDayIndex :: Day -> Opium.Connection -> IO (Either Opium.Error (Maybe DayIndex))
+readDayIndex :: Day -> Opium.Connection -> ExceptT Error IO (Maybe DayIndex)
 readDayIndex date =
-  Opium.fetch
-    "SELECT * FROM yore.day_index WHERE day = $1"
-    (Identity date)
+  lift
+    . Opium.fetch
+      "SELECT * FROM yore.day_index WHERE day = $1"
+      (Identity date)
 
-readDayPaths :: Day -> Opium.Connection -> IO (Either Opium.Error [(DayIndex, DayFile)])
+readDayPaths :: Day -> Opium.Connection -> ExceptT Error IO [(DayIndex, DayFile)]
 readDayPaths date =
-  Opium.fetch
-    "SELECT * FROM yore.day_index LEFT JOIN yore.day_file USING (day_index_id) WHERE day = $1"
-    (Identity date)
+  lift
+    . Opium.fetch
+      "SELECT * FROM yore.day_index LEFT JOIN yore.day_file USING (day_index_id) WHERE day = $1"
+      (Identity date)
 
-getDayFileByIssue :: Day -> Int -> Opium.Connection -> IO (Either Opium.Error DayFile)
+getDayFileByIssue :: Day -> Int -> Opium.Connection -> ExceptT Error IO DayFile
 getDayFileByIssue date issue =
   ex runIdentity
     . Opium.fetch
       "SELECT yore.day_file.* FROM yore.day_index LEFT JOIN yore.day_file USING (day_index_id) WHERE day = $1 ORDER BY day_file_id LIMIT 1 OFFSET $2"
       (date, issue)
 
-getNumberOfIssues :: Day -> Opium.Connection -> IO (Either Opium.Error Int)
+getNumberOfIssues :: Day -> Opium.Connection -> ExceptT Error IO Int
 getNumberOfIssues date =
   ex (count . runIdentity)
     . Opium.fetch
       "SELECT count(*) FROM yore.day_index LEFT JOIN yore.day_file USING (day_index_id) WHERE day = $1"
       (Identity date)
 
-ex :: (a -> b) -> IO (Either e a) -> IO (Either e b)
-ex = fmap . fmap
+ex :: (a -> b) -> IO (Either Opium.Error a) -> ExceptT Error IO b
+ex f = fmap f . lift

src/Yore/Download.hs

@@ -2,7 +2,8 @@
 
 module Yore.Download (downloadInto) where
 
-import Control.Monad (guard)
+import Control.Exception (Exception, IOException, catch)
+import Control.Monad.Trans.Except (ExceptT (..))
 import Data.Maybe (fromJust)
 import Data.Text (Text)
 import Network.HTTP.Req hiding (queryParam)
@@ -14,6 +15,8 @@ import qualified Data.Text as Text
 import qualified Data.UUID.V4 as UUID
 import qualified Text.URI as URI
 
+import Yore.Error (Error (..))
+
 -- | Download a URL and save it to a directory.
 -- Returns the path of the downloaded file relative to the directory.
 downloadInto
@@ -21,22 +24,30 @@ downloadInto
   -- ^ Directory where to store the file.
   -> Text
   -- ^ The URL to download.
-  -> IO FilePath
-downloadInto downloadDir textUrl = download >>= save
+  -> ExceptT Error IO FilePath
+downloadInto downloadDir textUrl =
+  download >>= save
   where
-    download :: IO LBS.ByteString
-    download = do
+    download :: ExceptT Error IO LBS.ByteString
+    download = toExceptT (handle @HttpException) $ do
       uri <- URI.mkURI textUrl
       let (url, opts) = fromJust $ useHttpsURI uri
       res <- runReq defaultHttpConfig $ req GET url NoReqBody lbsResponse opts
-      guard $ responseStatusCode res == 200
       pure $ responseBody res
 
-    save :: LBS.ByteString -> IO FilePath
-    save contents = do
+    save :: LBS.ByteString -> ExceptT Error IO FilePath
+    save contents = toExceptT (handle @IOException) $ do
       let suffix = takeExtension $ Text.unpack textUrl
       uuid <- UUID.nextRandom
       let fullPath = downloadDir </> show uuid <.> suffix
       createDirectoryIfMissing True downloadDir
       LBS.writeFile fullPath contents
       pure $ makeRelative downloadDir fullPath
+
+    toExceptT :: (Exception e) => (e -> Error) -> IO a -> ExceptT Error IO a
+    toExceptT showException f =
+      ExceptT $ (Right <$> f) `catch` (pure . Left . showException)
+
+    handle :: (Show e) => e -> Error
+    handle =
+      GenericError . Text.pack . show

src/Yore/Index.hs

@@ -1,24 +0,0 @@
-{-# LANGUAGE OverloadedRecordDot #-}
-{-# LANGUAGE OverloadedStrings #-}
-
-module Yore.Index (createEntries) where
-
-import Control.Exception (Exception, throwIO)
-import Control.Monad (forM_)
-import Data.Text (Text)
-import Data.Time (Day)
-
-import qualified Database.PostgreSQL.Opium as Opium
-
-import qualified Yore.DB as DB
-
-createEntries :: Day -> [(Text, FilePath)] -> Opium.Connection -> IO ()
-createEntries date urls conn = do
-  confidently $ Opium.execute_ "BEGIN" conn
-  dayIndex <- confidently $ DB.createDayIndex date conn
-  forM_ urls $ \(text, url) ->
-    confidently $ DB.createDayFile dayIndex.day_index_id text url conn
-  confidently $ Opium.execute_ "COMMIT" conn
-
-confidently :: (Exception e) => IO (Either e a) -> IO a
-confidently action = action >>= either throwIO pure

src/Yore/Log.hs

@@ -1,21 +1,22 @@
 module Yore.Log (Yore.Log.error, info) where
 
+import Control.Monad.IO.Class (MonadIO (..))
 import Data.Time (ZonedTime, getZonedTime)
 import Data.Time.Format (defaultTimeLocale, formatTime)
 import GHC.Stack (HasCallStack, SrcLoc (..), callStack, getCallStack)
 import Text.Printf (printf)
 
-info :: (HasCallStack) => String -> IO ()
+info :: (HasCallStack, MonadIO m) => String -> m ()
 info = doLog "INF"
 
-error :: (HasCallStack) => String -> IO ()
+error :: (HasCallStack, MonadIO m) => String -> m ()
 error = doLog "ERR"
 
-doLog :: (HasCallStack) => String -> String -> IO ()
-doLog level msg = do
+doLog :: (HasCallStack, MonadIO m) => String -> String -> m ()
+doLog level msg = liftIO $ do
   now <- getZonedTime
   let location = getLocation $ getCallStack callStack
-  printf "(%s) (%s) (%s) %s\n" (iso8601Show now) location level msg
+  printf "(%s) (%s) (%s) %s\n" (iso8601Show now) level location msg
   where
     getLocation :: [(String, SrcLoc)] -> String
     -- First entry is always a function from this module, skip it

src/Yore/Repl.hs

@@ -1,59 +0,0 @@
-{-# LANGUAGE LambdaCase #-}
-{-# LANGUAGE OverloadedRecordDot #-}
-{-# LANGUAGE OverloadedStrings #-}
-
-module Yore.Repl (connect, exec, getToday, indexDay, addYears) where
-
-import Control.Exception (bracket, throwIO)
-import Control.Monad (forM)
-import Data.Time (Day, getZonedTime, localDay, zonedTimeToLocalTime)
-import Text.Printf (printf)
-
-import qualified Database.PostgreSQL.Opium as Opium
-
-import Yore.DB (DayIndex)
-import Yore.Download (downloadInto)
-import Yore.Index (createEntries)
-import Yore.Scrape (Issue (..), getIssuesByDay)
-import Yore.Time (addYears)
-
-import qualified Yore.DB as DB
-
-connect :: IO Opium.Connection
-connect =
-  Opium.connect "host=localhost port=5433 user=yore-test" >>= \case
-    Left e -> error $ show e
-    Right c -> pure c
-
-exec :: (Opium.Connection -> IO a) -> IO a
-exec = bracket connect Opium.close
-
-getToday :: IO [Issue]
-getToday = do
-  now <- getZonedTime
-  print now
-  let currentDay = localDay $ zonedTimeToLocalTime now
-  case addYears (-100) currentDay of
-    Right then_ ->
-      getIssuesByDay then_
-    Left _ ->
-      pure []
-
-indexDay :: Day -> IO ()
-indexDay date = exec $ \conn -> do
-  either throwIO pure =<< Opium.execute_ "SELECT pg_advisory_lock(42);" conn
-  printf "Checking whether %s already exists...\n" (show date)
-  mbDi <- either throwIO pure =<< DB.readDayIndex date conn
-  case mbDi of
-    Just (_ :: DayIndex) ->
-      putStrLn "Nothing to do."
-    Nothing -> do
-      putStrLn "Retrieving URLs..."
-      issues <- getIssuesByDay date
-      paths <- forM issues $ \issue -> do
-        printf "Downloading %s...\n" issue.url
-        path <- downloadInto "./download" issue.url
-        pure (issue.label, path)
-      putStrLn "Creating DB entries..."
-      createEntries date paths conn
-      putStrLn "Done."

src/Yore/Schedule.hs

@@ -0,0 +1,22 @@
+module Yore.Schedule (schedule) where
+
+import Control.Concurrent (forkIO, threadDelay)
+import Control.Monad (when)
+import Data.Functor (($>))
+import Data.Time (ZonedTime, getZonedTime, nominalDiffTimeToSeconds, secondsToNominalDiffTime)
+import Data.Time.Clock.POSIX (getPOSIXTime)
+
+schedule :: (ZonedTime -> Bool) -> IO a -> IO ()
+schedule shouldRunAt f = everyMinute $ \now -> when (shouldRunAt now) (f $> ())
+
+everyMinute :: (ZonedTime -> IO ()) -> IO ()
+everyMinute f = do
+  -- Use POSIX time to avoid having to handle leap seconds
+  now <- getPOSIXTime
+  let posixSeconds :: Int
+      posixSeconds = floor $ nominalDiffTimeToSeconds now
+      lastFullMinute = secondsToNominalDiffTime $ fromIntegral $ posixSeconds - posixSeconds `mod` 60
+      secondsSinceLastFullMinute = now - lastFullMinute
+  threadDelay $ ceiling $ (60 - secondsSinceLastFullMinute) * 1_000_000
+  _ <- forkIO $ f =<< getZonedTime
+  everyMinute f

src/Yore/Servant.hs

@@ -0,0 +1,38 @@
+{-# LANGUAGE OverloadedRecordDot #-}
+{-# LANGUAGE OverloadedStrings #-}
+{-# LANGUAGE TypeFamilies #-}
+
+module Yore.Servant (GetSendfile, Sendfile (..), HTML) where
+
+import Network.HTTP.Media ((//), (/:))
+import Network.HTTP.Types (ResponseHeaders, status200)
+import Network.Wai (responseFile)
+import Servant (Accept (..), HasServer (..))
+import Servant.Server.Internal.Delayed (runAction)
+import Servant.Server.Internal.RouteResult (RouteResult (..))
+import Servant.Server.Internal.Router (Router' (..))
+
+-- | Servant route type that responds with a file using @sendfile@.
+data GetSendfile
+
+data Sendfile = Sendfile
+  { headers :: ResponseHeaders
+  -- ^ The headers to send along with the file, e.g. @content-type@ or cache information.
+  , path :: FilePath
+  -- ^ The file to respond with.
+  }
+  deriving (Show)
+
+instance HasServer GetSendfile context where
+  type ServerT GetSendfile m = m Sendfile
+
+  hoistServerWithContext _ _ nt = nt
+
+  route _ _ action = RawRouter $ \env request respond ->
+    runAction action env request respond $ \sendfile ->
+      Route $ responseFile status200 sendfile.headers sendfile.path Nothing
+
+data HTML
+
+instance Accept HTML where
+  contentType _ = "text" // "html" /: ("charset", "utf-8")

yore.cabal

@@ -1,7 +1,7 @@
 cabal-version: 3.4
 
 name:           yore
-version:        0
+version:        0.0.6
 author:         Paul Brinkmeier
 maintainer:     hallo@pbrinkmeier.de
 copyright:      2023 Paul Brinkmeier
@@ -37,11 +37,11 @@ library
     , Yore.DB
     , Yore.Download
     , Yore.Error
-    , Yore.Index
     , Yore.Log
-    , Yore.Repl
+    , Yore.Schedule
     , Yore.Scrape
     , Yore.Time
+    , Yore.Servant
   hs-source-dirs:
       src
   build-depends:
@@ -50,16 +50,20 @@ library
     , directory
     , filepath
     , html-parse
+    , http-media
+    , http-types
     , lens
     , modern-uri
     , mtl
     , opium
     , req
     , resource-pool
+    , servant-server
     , text
     , time
     , transformers
     , uuid
+    , wai
 
 executable yore
   import: shared-options
@@ -78,7 +82,6 @@ executable yore
     , bytestring
     , filepath
     , opium
-    , http-media
     , http-types
     , servant-server
     , text