Report geography no dataset declared

CHANGELOG.md

@@ -3,6 +3,16 @@
 ## [Unreleased]
 
 ### Added
+- The quality report now says when a geography was never declared. Nearly every
+  process of some SimaPro databases leaves the `Geography` field at
+  `Unspecified` — 97% of Agribalyse 3.2 — and the only geography left is the
+  code inside the dataset name (`… {FR}`, `…//[RER]`, `…/CN U`). VoLCA reads
+  it, which is what makes those databases usable, but the result is a reading of
+  a name and not a declaration, and downstream the two are the same text. An
+  EcoSpold dataset with no geography at all is filled in with `GLO`, likewise
+  silently. The new `undeclaredGeography` check counts both, one finding per
+  entry, so a maintainer can see how much of a database's geography is source
+  data before treating it as such.
 - A warning now reports factor lines that match a flow but cannot be
   converted into its unit — a per-kilogram factor against a flow measured in
   cubic metres, for instance. Refusing such a factor is correct (the

src/API/DatabaseHandlers.hs

@@ -342,6 +342,7 @@ qualityReportToAPI mLimit r =
         , qraDuplicateActivities = checkToAPI mLimit (Quality.qrDuplicateActivities r)
         , qraSuspiciousAmounts = checkToAPI mLimit (Quality.qrSuspiciousAmounts r)
         , qraMissingMetadata = checkToAPI mLimit (Quality.qrMissingMetadata r)
+        , qraUndeclaredGeography = checkToAPI mLimit (Quality.qrUndeclaredGeography r)
         , qraFormulaConsistency = checkToAPI mLimit (Quality.qrFormulaConsistency r)
         , qraTruncatedNameCollisions = checkToAPI mLimit (Quality.qrTruncatedNameCollisions r)
         , qraMissingPedigree = checkToAPI mLimit (Quality.qrMissingPedigree r)

src/API/Resources.hs

@@ -478,7 +478,10 @@ description r = case r of
         \coproducts carry one), entries duplicated outright \
         \(same name, location and reference product), non-finite amounts or a \
         \zero reference amount, missing metadata (description, \
-        \classification, location, units absent from the registry), stored \
+        \classification, location, units absent from the registry), \
+        \geography the source never declared — read off the dataset name \
+        \(SimaPro writes 'Unspecified' in whole databases) or filled in by the \
+        \loader — stored \
         \amounts that disagree with the formulas documenting them \
         \(mathematicalRelation, checked at parse time), distinct names \
         \that merge under SimaPro's 80-character truncation, exchanges \

src/API/Types.hs

@@ -847,6 +847,7 @@ data QualityReportAPI = QualityReportAPI
     , qraDuplicateActivities :: QualityCheckAPI
     , qraSuspiciousAmounts :: QualityCheckAPI
     , qraMissingMetadata :: QualityCheckAPI
+    , qraUndeclaredGeography :: QualityCheckAPI
     , qraFormulaConsistency :: QualityCheckAPI
     , qraTruncatedNameCollisions :: QualityCheckAPI
     , qraMissingPedigree :: QualityCheckAPI

src/BrightwayExcel/Parser.hs

@@ -321,13 +321,15 @@ rawToActivity cfg ra =
                | not (any exchangeIsReference exchanges')
                ]
 
+    location = fromMaybe "" (metaText meta "location")
     activity =
         Activity
             { activityName = raName ra
             , activityDescription = maybeToList (metaText meta "comment")
             , activitySynonyms = M.empty
             , activityClassification = M.empty
-            , activityLocation = fromMaybe "" (metaText meta "location")
+            , activityLocation = location
+            , activityLocationSource = declaredLocationSource location
             , activityUnit = refUnitName
             , exchanges = exchanges'
             , activityParams = M.empty

src/Database/Loader.hs

@@ -216,14 +216,17 @@ History of manual bumps:
      when that name carries one, instead of always being minted from
      name and location. Old caches key the same dataset under the minted
      UUID, so a mixed pair would compare as two disjoint databases.
+- 13: Activity record gained activityLocationSource (declared, read off the
+     dataset name, or neither). Old caches miss the field; the Store layout
+     is positional, so decoding them would misread every field after it.
 
 The signature is stored inside the cache file and checked on load.
 If it doesn't match, the cache is automatically invalidated and rebuilt.
 -}
 schemaSignature :: Word64
 schemaSignature =
     let Fingerprint hi lo = typeRepFingerprint (typeRep (Proxy :: Proxy Database))
-     in hi `xor` lo `xor` 12
+     in hi `xor` lo `xor` 13
 
 {- |
 Helper function to parse UUID from Text with deterministic UUID generation fallback.

src/Database/Quality.hs

@@ -10,10 +10,11 @@ amounts that aren't finite, missing metadata, stored amounts that disagree
 with the formulas documenting them, distinct names that merge under
 SimaPro's 80-character truncation, exchanges without the pedigree scores
 their database otherwise carries, reference products nothing in the
-database consumes, land transformation whose "to" and "from" areas
-don't balance within an activity, oxygen-demand or organic-carbon
-measures reported in a physically impossible order, CAS numbers
-whose check digit doesn't confirm them, allocation percentages
+database consumes, geography no dataset declared (read off the name
+or filled in by the loader instead), land transformation whose "to" and
+"from" areas don't balance within an activity, oxygen-demand or
+organic-carbon measures reported in a physically impossible order, CAS
+numbers whose check digit doesn't confirm them, allocation percentages
 outside the 0-100% range, and amounts too small to have been measured.
 
 Every check is a pure scan over a 'SimpleDatabase', so the report is
@@ -46,6 +47,7 @@ import Types (
     BiosphereFlow (..),
     Exchange (..),
     FormulaCheck (..),
+    LocationSource (..),
     Severity (..),
     SimpleDatabase (..),
     TechRole (..),
@@ -102,6 +104,7 @@ data QualityReport = QualityReport
     , qrDuplicateActivities :: !QualityCheck
     , qrSuspiciousAmounts :: !QualityCheck
     , qrMissingMetadata :: !QualityCheck
+    , qrUndeclaredGeography :: !QualityCheck
     , qrFormulaConsistency :: !QualityCheck
     , qrTruncatedNameCollisions :: !QualityCheck
     , qrMissingPedigree :: !QualityCheck
@@ -125,6 +128,7 @@ qualityChecks r =
     , qrDuplicateActivities r
     , qrSuspiciousAmounts r
     , qrMissingMetadata r
+    , qrUndeclaredGeography r
     , qrFormulaConsistency r
     , qrTruncatedNameCollisions r
     , qrMissingPedigree r
@@ -225,6 +229,7 @@ qualityReport dbName db =
         , qrDuplicateActivities = QualityCheck True (worstFirst duplicateOffenders)
         , qrSuspiciousAmounts = QualityCheck True (worstFirst amountOffenders)
         , qrMissingMetadata = QualityCheck True (worstFirst metadataOffenders)
+        , qrUndeclaredGeography = QualityCheck True (worstFirst geographyOffenders)
         , qrFormulaConsistency = QualityCheck formulaApplicable (worstFirst formulaOffenders)
         , qrTruncatedNameCollisions = QualityCheck True (worstFirst truncationOffenders)
         , qrMissingPedigree = QualityCheck pedigreeApplicable (worstFirst pedigreeOffenders)
@@ -435,6 +440,27 @@ qualityReport dbName db =
             | (key, act) <- entries
             ]
 
+    -- A geography the source declares is a fact about the dataset; one read off
+    -- the dataset name is this loader's reading of a string, and one supplied by
+    -- the loader is neither. SimaPro writes "Unspecified" in the Geography field
+    -- of entire databases, so their geography survives only in names like
+    -- "… {FR}"; an EcoSpold dataset with no geography is filled in with "GLO".
+    -- Both are usable and neither was declared, and downstream the two are the
+    -- same text — hence Info, and hence a count: a maker reads it before
+    -- treating the geography as source data.
+    geographyOffenders =
+        [ offender InfoSev key act Nothing detail
+        | (key, act) <- entries
+        , Just detail <- [undeclaredGeography (activityLocationSource act) (activityLocation act)]
+        ]
+    undeclaredGeography src loc = case src of
+        LocationDeclared -> Nothing
+        LocationInferredFromName ->
+            Just ("geography \"" <> loc <> "\" was read from the dataset name, not declared by the source")
+        LocationUnspecified
+            | T.null (T.strip loc) -> Just "the source declares no geography"
+            | otherwise -> Just ("the source declares no geography; \"" <> loc <> "\" stands in for it")
+
     -- Names that only differ beyond SimaPro's cap become one name on export —
     -- the over-grouping the allocation check above tolerates at parse time
     -- turns into data loss on the way out. One finding per distinct name, each

src/EcoSpold/Parser1.hs

@@ -453,14 +453,17 @@ buildResult :: ParseState -> Either String (Activity, [TechnosphereFlow], [Biosp
 buildResult st =
     let name = fromMaybe "Unknown Activity" (psActivityName st)
         location = fromMaybe "GLO" (psLocation st)
+        -- "GLO" above is this loader's stand-in, not something the dataset said:
+        -- a dataset without a geography is recorded as declaring none.
+        locationSource = maybe LocationUnspecified declaredLocationSource (psLocation st)
         refUnit = fromMaybe "UNKNOWN_UNIT" (psRefUnit st)
         description = reverse (psDescription st)
         classifications =
             M.fromList $
                 filter
                     (not . T.null . snd)
                     [("Category", psActivityCategory st), ("SubCategory", psActivitySubCategory st)]
-        activity = Activity name description M.empty classifications location refUnit (reverse $ psExchanges st) M.empty M.empty Nothing Nothing Nothing Nothing Nothing
+        activity = Activity name description M.empty classifications location locationSource refUnit (reverse $ psExchanges st) M.empty M.empty Nothing Nothing Nothing Nothing Nothing
         pack act =
             ( act
             , reverse (psTechFlows st)

src/EcoSpold/Parser2.hs

@@ -776,13 +776,16 @@ parseWithXeno xmlContent processId = do
     buildResult st _pid =
         let name = fromMaybe "Unknown Activity" (psActivityName st)
             location = fromMaybe "GLO" (psLocation st)
+            -- "GLO" above is this loader's stand-in, not something the dataset
+            -- said: a dataset without a geography declares none.
+            locationSource = maybe LocationUnspecified declaredLocationSource (psLocation st)
             description = reverse (psDescription st) -- Reverse to get correct order
             refUnit = fromMaybe "UNKNOWN_UNIT" (psRefUnit st)
             nativeType = ecoSpoldNativeType (psActivityType st) (psSpecialActivityType st)
             pairs = reverse (psExchanges st)
             formulaCheck = checkFormulas (psParams st) pairs
             -- Apply cutoff strategy to exchanges
-            activity = Activity name description M.empty (psClassifications st) location refUnit (map fst pairs) (psParams st) (psParamExprs st) Nothing Nothing nativeType Nothing formulaCheck
+            activity = Activity name description M.empty (psClassifications st) location locationSource refUnit (map fst pairs) (psParams st) (psParamExprs st) Nothing Nothing nativeType Nothing formulaCheck
             techs = reverse (psTechFlows st)
             bios = reverse (psBioFlows st)
             wastes = reverse (psWasteFlows st)

src/ILCD/Parser.hs

@@ -560,6 +560,7 @@ buildActivity flowInfoMap techFlowDB bioFlowDB wasteFlowDB unitDB p =
         , activitySynonyms = M.empty
         , activityClassification = iprClassifications p
         , activityLocation = iprLocation p
+        , activityLocationSource = declaredLocationSource (iprLocation p)
         , activityUnit = refUnit
         , exchanges = map (mkExchange (iprRefFlowIdx p)) (iprExchanges p)
         , activityParams = M.empty

src/SimaPro/Parser.hs

@@ -19,7 +19,7 @@ module SimaPro.Parser (
     normalizeSimaProCompartment,
     extractLocation,
     Located (..),
-    LocationSource (..),
+    NameReading (..),
 
     -- * Shared utilities (used by Method.ParserSimaPro)
     defaultConfig,
@@ -791,14 +791,16 @@ generateUnitUUID unitName =
 -- Conversion to volca Types
 -- ============================================================================
 
-{- | How a location was read out of a SimaPro name.
+{- | How a location was read out of a SimaPro name. (Not to be confused with
+'Types.LocationSource', which says whether an activity's location was declared
+at all — every reading here is an inferred one by that measure.)
 
 A tag is a location the producer wrote down. A slash suffix is a guess made by
 cutting the name, and names end in a slash for reasons that have nothing to do
 with geography — in "Already packed - PP/PE", PE is a plastic, not Peru. So the
 ordering here says a tag beats a suffix, wherever each of the two sits.
 -}
-data LocationSource
+data NameReading
     = Tagged
     | SlashSuffix
     deriving (Eq, Ord, Show)
@@ -810,7 +812,7 @@ data Located = Located
     informational, shortened for a suffix, which is part of the name.
     -}
     , locatedLocation :: Text
-    , locatedSource :: LocationSource
+    , locatedSource :: NameReading
     }
     deriving (Eq, Show)
 
@@ -931,7 +933,8 @@ processBlockToActivity unitCfg GlobalParams{..} ProcessBlock{..} =
     processReading = extractLocation pbName
 
     -- The Geography field, when the producer filled it in. It outranks anything
-    -- read out of a name, being the one place meant to hold a location.
+    -- read out of a name, being the one place meant to hold a location — a name
+    -- reading is recorded as such below, for the quality report.
     statedLocation = mfilter ((/= "unspecified") . T.toLower) (nonEmptyText pbLocation)
 
     {- Trimmed Process name (without curly-brace location tag). Empty when the
@@ -1013,6 +1016,10 @@ processBlockToActivity unitCfg GlobalParams{..} ProcessBlock{..} =
                     . sortOn locatedSource
                     $ catMaybes readings
             effectiveLoc = fromMaybe readLocation statedLocation
+            effectiveLocSource
+                | isJust statedLocation = LocationDeclared
+                | T.null effectiveLoc = LocationUnspecified
+                | otherwise = LocationInferredFromName
             allocFormula = mfilter (not . isNumericFormula) (nonEmptyText (prAllocRaw prod))
             activity =
                 Activity
@@ -1027,6 +1034,7 @@ processBlockToActivity unitCfg GlobalParams{..} ProcessBlock{..} =
                                 , ("Category", prCategory prod)
                                 ]
                     , activityLocation = effectiveLoc
+                    , activityLocationSource = effectiveLocSource
                     , activityUnit = effUnitName
                     , exchanges = productExchange : map (scaleExchange allocFraction) sharedExchanges
                     , activityParams = env

src/Types.hs

@@ -483,6 +483,31 @@ data FormulaCheck = FormulaCheck
     }
     deriving (Generic, NFData, Store)
 
+{- | Where an activity's 'activityLocation' came from. A format with a geography
+field publishes one ('LocationDeclared'); when it is blank or says
+@Unspecified@, the parser falls back to the code embedded in the dataset name
+(@"… {FR}"@, @"…//[RER]"@, @"…/CN U"@), which is a reading of the name and not
+a declaration ('LocationInferredFromName'). 'LocationUnspecified' when neither
+yielded anything, and 'activityLocation' is empty.
+
+Recorded at parse time because only the parser can still tell the two apart:
+downstream, a declared @FR@ and a guessed @FR@ are the same text. The quality
+report is what surfaces the difference.
+-}
+data LocationSource
+    = LocationDeclared
+    | LocationInferredFromName
+    | LocationUnspecified
+    deriving (Show, Eq, Generic, NFData, Store)
+
+{- | The source of a geography a format declares directly: the field itself when
+it carries something, nothing to infer from when it doesn't.
+-}
+declaredLocationSource :: Text -> LocationSource
+declaredLocationSource loc
+    | T.null (T.strip loc) = LocationUnspecified
+    | otherwise = LocationDeclared
+
 {- | Base LCA activity
 Note: ProcessId is the index in dbActivities vector, UUIDs stored in dbProcessIdTable
 -}
@@ -492,6 +517,7 @@ data Activity = Activity
     , activitySynonyms :: !(M.Map Text (S.Set Text)) -- Synonyms by language, same structure as flows
     , activityClassification :: !(M.Map Text Text) -- Classifications (ISIC, CPC, etc.)
     , activityLocation :: !Text -- Location code (e.g. FR, RER)
+    , activityLocationSource :: !LocationSource -- Whether the source declared that location or the parser read it off the dataset name
     , activityUnit :: !Text -- Reference unit
     , exchanges :: ![Exchange] -- List of exchanges
     , activityParams :: !(M.Map Text Double) -- Resolved dataset parameter values (SimaPro parameters, EcoSpold2 <parameter> variables)

test/BM25Spec.hs

@@ -29,6 +29,7 @@ mkActivity name loc xs =
         , activitySynonyms = M.empty
         , activityClassification = M.empty
         , activityLocation = loc
+        , activityLocationSource = LocationDeclared
         , activityUnit = "kg"
         , exchanges = xs
         , activityParams = M.empty

test/BrightwayExcelWriterSpec.hs

@@ -261,6 +261,7 @@ elec =
         , activitySynonyms = M.empty
         , activityClassification = M.empty
         , activityLocation = "GLO"
+        , activityLocationSource = LocationDeclared
         , activityUnit = "kilowatt hour"
         , exchanges = [prodExch, gasExch, co2Exch]
         , activityParams = M.empty

test/CopySpec.hs

@@ -44,6 +44,7 @@ import Types (
     Database (..),
     Exchange (..),
     GeographyPolicy (..),
+    LocationSource (..),
     SparseTriple (..),
     TechRole (..),
     TechnosphereFlow (..),
@@ -274,6 +275,7 @@ supplierDB offset products =
                         , activitySynonyms = M.empty
                         , activityClassification = M.empty
                         , activityLocation = "GLO"
+                        , activityLocationSource = LocationDeclared
                         , activityUnit = "kg"
                         , exchanges = [refOut]
                         , activityParams = M.empty

test/CrossDBActivityLinkSpec.hs

@@ -69,6 +69,7 @@ mkActivity name loc exs =
         , activitySynonyms = M.empty
         , activityClassification = M.empty
         , activityLocation = loc
+        , activityLocationSource = LocationDeclared
         , activityUnit = "kg"
         , exchanges = exs
         , activityParams = M.empty

test/CrossDBRegionalLCIAFixture.hs

@@ -103,6 +103,7 @@ mkActivity _ loc =
         , activitySynonyms = M.empty
         , activityClassification = M.empty
         , activityLocation = loc
+        , activityLocationSource = LocationDeclared
         , activityUnit = "kg"
         , exchanges = []
         , activityParams = M.empty

test/CrossLinkingSpec.hs

@@ -445,6 +445,7 @@ mkActivityAt loc =
         , activitySynonyms = M.empty
         , activityClassification = M.empty
         , activityLocation = loc
+        , activityLocationSource = LocationDeclared
         , activityUnit = "kg"
         , exchanges = []
         , activityParams = M.empty

test/DeleteSelectionSpec.hs

@@ -51,6 +51,7 @@ import Types (
     Database (..),
     Exchange (..),
     GeographyPolicy (..),
+    LocationSource (..),
     ProcessId,
     SparseTriple (..),
     TechRole (..),
@@ -501,6 +502,7 @@ mkActivity name loc classif exs =
         , activitySynonyms = M.empty
         , activityClassification = classif
         , activityLocation = loc
+        , activityLocationSource = LocationDeclared
         , activityUnit = "kg"
         , exchanges = exs
         , activityParams = M.empty

test/EcoSpold1Spec.hs

@@ -25,6 +25,7 @@ emptyActivity =
         , activitySynonyms = M.empty
         , activityClassification = M.empty
         , activityLocation = "GLO"
+        , activityLocationSource = LocationDeclared
         , activityUnit = "kg"
         , exchanges = []
         , activityParams = M.empty