feat: file each commodity row in its Schedule B unit, and record Incoterms on the CEVA SLI

README.md

@@ -57,9 +57,12 @@ of defaulting to zero.
 6. **Validates every Schedule B number** against the U.S. Census Bureau AES commodity file:
    ten digits, currently active, reported in the required unit of quantity, and plausibly
    describing the goods.
-7. **Proves the arithmetic** before generating anything. Quantities, weights and values must
+7. **Files each row in the unit its commodity number is reported in.** Where Schedule B
+   reports a code by weight and the invoice counts pieces, the row carries the net weight
+   instead — and the review screen lets you change the unit before anything is generated.
+8. **Proves the arithmetic** before generating anything. Quantities, weights and values must
    sum back to the totals printed on the source document.
-8. **Fills the carrier's real blank form** and downloads it, still editable and unsigned.
+9. **Fills the carrier's real blank form** and downloads it, still editable and unsigned.
 
 ## Supported carriers
 
@@ -282,7 +285,33 @@ This is desktop-only because it has to be: census.gov serves the file with no
 The dataset is the authority on three things the CIPL cannot tell you: whether a code is
 currently valid, its official description, and the unit of quantity AES requires. That last
 one matters more than it looks — `9031.90.0000` and `8483.10.5000` are reported in
-kilograms, not pieces, and the tool flags a piece count filed against them.
+kilograms, not pieces, and a piece count filed against them is a reporting error even when
+every other number on the form is right.
+
+## Unit of quantity
+
+Each commodity row is filed in the unit its Schedule B number requires, wherever the
+shipment can state it. A code reported in kilograms carries the row's net weight rather than
+the invoice's piece count; a code reported by the piece is unchanged. Both carrier forms and
+both keying sheets take the same figure from the same place, so the paperwork prepared for
+one shipment cannot disagree with itself.
+
+Where a code accepts more than one unit — `NO+KG` — only the filer knows how the goods are
+actually measured, so the review screen offers the choice per commodity number. Units the
+shipment cannot state are listed and disabled with the reason, rather than hidden: "there is
+no net weight for these goods" is the answer somebody is looking for. Where nothing supports
+the required unit at all, the document's own figure is filed and the reconciliation says so —
+no quantity is ever invented to fit a unit.
+
+## Incoterms
+
+The rule is read off whatever the document or the operator supplied, not matched literally:
+`DAP Singapore`, `FOB Origin - Collect`, `cif rotterdam` and `Ex Works` all resolve to the
+rule the form has a box for. The named place is kept — an Incoterm without one does not say
+which port — and written into the CEVA form's special instructions, which is the only place
+that form can record it. A rule withdrawn since Incoterms 2020 (`DAT`, `DDU`) is reported
+with its replacement rather than silently mapped onto it: reclassifying a delivery term is
+the filer's decision.
 
 ## Local data
 

roadmap.md

@@ -27,6 +27,8 @@ B refresh, which cannot work in a browser.
 | **Second CIPL format** | ✅ Done | The SAP-style `SHIPMENT#` layout behind a format registry: single currency, stated ECCNs, no weights, two-digit dates. |
 | **Reconciliation** | ✅ Done | USD set selection, invoice↔packing-list join, grouping by Schedule B + D/F, blocking totals checks. |
 | **Schedule B validation** | ✅ Done | Census AES dataset: 10 digits, currently active, required unit of quantity, description plausibility. Staleness warning once a January/July revision passes the dataset's date. |
+| **Unit of quantity** | ✅ Done | Each row filed in the unit its commodity number requires — the net weight where Schedule B reports a code by weight — with the unit changeable per code on the review screen, carried identically into both SLIs and both keying sheets. Never invents a figure a shipment cannot state. |
+| **Incoterms** | ✅ Done | The rule read off `DAP Singapore`, `FOB Origin - Collect` or `Ex Works` rather than matched literally, with the named place preserved and written where a form has no box for it. Rules withdrawn since Incoterms 2020 reported, never remapped. |
 | **Item library** | ✅ Done | Item-master import (.xlsx/.csv/.tsv, columns matched by heading, explicit weight unit) supplying per-part weights, with the `####.##.####` + concordance filter flagging codes to fix by part number. |
 | **Nippon Express adapter** | ✅ Done | Blank form 01/04/2022, 8 commodity rows, per-cell fields. |
 | **CEVA adapter** | ✅ Done | Blank form 11201-C3 rev. 8/2023, multiline column fields, whole-dollar values. |

src/App.tsx

@@ -74,6 +74,14 @@ export function App() {
   const [carrierId, setCarrierId] = useState<ShipmentCarrierId>('nippon-express')
   const [profile, setProfile] = useState<CompanyProfile>(EMPTY_PROFILE)
   const [settings, setSettings] = useState<ShipmentSettings>(() => defaultShipmentSettings(getAdapter('nippon-express')))
+  /**
+   * The unit of quantity to file each commodity number in, keyed by normalised code.
+   *
+   * Per shipment rather than saved, and empty unless somebody changes something: the default
+   * already follows Schedule B wherever these documents can state it, and a unit remembered
+   * from a previous shipment would quietly file goods that were measured differently.
+   */
+  const [reportingUnits, setReportingUnits] = useState<Record<string, string>>({})
   const [overrides, setOverrides] = useState<OverrideRecord[]>([])
   const [partOverrides, setPartOverrides] = useState<PartOverrideRecord[]>([])
   const [items, setItems] = useState<ItemLibraryEntry[]>([])
@@ -255,6 +263,10 @@ export function App() {
       setBusy(true)
       setParsed(next)
       setError(null)
+      // A unit was chosen for the commodity numbers on the *last* document. These goods are
+      // measured however this document says they are, so the choice starts again from the
+      // Schedule B default rather than carrying over onto codes that happen to repeat.
+      setReportingUnits({})
 
       try {
         const header = next.headers[next.availableSets[0]]
@@ -338,9 +350,10 @@ export function App() {
       // Only consulted for formats that state no weights; a printed weight always wins.
       unitWeightsByPart,
       itemsByPart,
+      reportingUnits,
       maxRows: adapter.maxCommodityRows,
     })
-  }, [parsed, scheduleB, settings.eccn, settings.sme, settings.license, overrides, codesByPart, unitWeightsByPart, itemsByPart, adapter])
+  }, [parsed, scheduleB, settings.eccn, settings.sme, settings.license, overrides, codesByPart, unitWeightsByPart, itemsByPart, reportingUnits, adapter])
 
   const draft = useMemo(
     () => (reconciliation ? buildDraft(reconciliation, profile, settings, adapter) : null),
@@ -829,7 +842,21 @@ export function App() {
                 onSaveDescription={(part, description, text) => void savePartDescription(part, description, text)}
                 onClearCode={(part) => void clearPartCode(part)}
               />
-              <CommodityTable reconciliation={reconciliation} />
+              <CommodityTable
+                reconciliation={reconciliation}
+                reportingUnits={reportingUnits}
+                onReportingUnitChange={(code, unit) =>
+                  setReportingUnits((current) => {
+                    // An empty choice is "use the Schedule B default", which is the absence of
+                    // an entry rather than an entry saying nothing — otherwise the blank sits
+                    // in the record and reads later as a unit somebody picked.
+                    const next = { ...current }
+                    if (unit) next[code] = unit
+                    else delete next[code]
+                    return next
+                  })
+                }
+              />
               <ChecksPanel checks={checks} canGenerate={canGenerate} />
               <OverridesPanel
                 reconciliation={reconciliation}