@willcgage/module-schematic 0.77.0 → 0.79.0

package/dist/index.d.cts

@@ -1486,6 +1486,15 @@ interface TrackPart {
     frogOffset?: PartDimension;
     /** End-to-end length of the part. */
     overallLength?: PartDimension;
+    /** Frog apex → the end of the diverging rail, measured ALONG that rail. The
+     * independent cross-check on {@link frogOffset}: it must be slightly LONGER
+     * than the axial `overallLength − frogOffset`, because the rail is the
+     * hypotenuse of the angle it leaves at. A reading that makes that difference
+     * negative means a mis-read frog — which is exactly how the 2057 wye's frog
+     * was caught (see its note). Too insensitive to recover the angle from — at
+     * these angles ¹⁄₃₂″ of slop swings the implied half-angle by 6° — so use it
+     * to falsify a frog position, never to derive one. */
+    divergingLength?: PartDimension;
     /** Diverging route radius (straight turnouts). */
     divergingRadius?: PartDimension;
     /** Curved turnouts: the two concentric radii. */
@@ -1520,6 +1529,15 @@ declare const CODE55_RAIL_HEIGHT_INCHES = 0.055;
  *     #5     1¾″      4.75″     6.00″     3.00000″   1.25″       0.2500
  *     #7     ⁵⁄₈″     4⁷⁄₃₂″    6.00″     3.59375″   1.78125″    0.2545
  *     #10    ⁹⁄₁₆″    5.50″     8.00″     4.93750″   2.50″       0.2500
+ *     wye    points   frog V    overall   lead       past frog   diverging rail
+ *     #2.5   1⁵⁄₈″    4⅛″       6.50″     2.50000″   2.375″      2.4375″
+ *     #3.5   ¾″       3⁵⁄₃₂″    5.00″     2.40625″   1.84375″    1.9375″
+ *
+ * ⚠️ THE WYES ARE NOT ON THE ÷N TREND AND MUST NOT BE INTERPOLATED WITH THE
+ * TURNOUTS. lead ÷ N is 1.00 and 0.69 against the turnouts' flat 0.25, and the
+ * #2.5 is the LONGER part despite the sharper frog. Every size lookup here
+ * filters `kind === "turnout"`, which is what keeps them out — do not "fix"
+ * that filter to include wyes.
  *
  * Every one of these is Will Gage's, off a physical part, frog taken at the V.
  * ⚠️ The #7 lead SUPERSEDES Steve Branton's 3⅜″ (#173) — the library's founding
@@ -1883,6 +1901,8 @@ interface StoredTrackPart {
     frogOffsetSource?: string | null;
     overallLengthInches?: number | null;
     overallLengthSource?: string | null;
+    divergingLengthInches?: number | null;
+    divergingLengthSource?: string | null;
     leadInches?: number | null;
     leadSource?: string | null;
     outerRadiusInches?: number | null;
@@ -1916,6 +1936,17 @@ declare function storedPartToTrackPart(row: StoredTrackPart): TrackPart;
  *
  * The built-ins remain the floor: a part nobody has stored still resolves, so
  * geometry keeps working with no database at all.
+ *
+ * ⛔ THE TRAP THIS CREATES, and it has already been sprung once. Replacement is
+ * WHOLESALE, so a stored row that is merely INCOMPLETE deletes every dimension
+ * the built-in had. When both Atlas wyes were measured in 0.78.0 their stored
+ * rows still held null offsets from the original seed — so in production the
+ * new measurements were invisible and the 2056 still carried a `derived` lead
+ * the release had just retired. Nothing failed; the parts simply had no
+ * dimensions, which is indistinguishable from never having measured them.
+ *
+ * **Measuring a built-in is therefore not finished until the stored row is
+ * updated too.** Adding a dimension here and stopping ships nothing.
  */
 declare function mergeStoredParts(stored: StoredTrackPart[], library?: TrackPart[]): TrackPart[];
 /**

package/dist/index.d.ts

@@ -1486,6 +1486,15 @@ interface TrackPart {
     frogOffset?: PartDimension;
     /** End-to-end length of the part. */
     overallLength?: PartDimension;
+    /** Frog apex → the end of the diverging rail, measured ALONG that rail. The
+     * independent cross-check on {@link frogOffset}: it must be slightly LONGER
+     * than the axial `overallLength − frogOffset`, because the rail is the
+     * hypotenuse of the angle it leaves at. A reading that makes that difference
+     * negative means a mis-read frog — which is exactly how the 2057 wye's frog
+     * was caught (see its note). Too insensitive to recover the angle from — at
+     * these angles ¹⁄₃₂″ of slop swings the implied half-angle by 6° — so use it
+     * to falsify a frog position, never to derive one. */
+    divergingLength?: PartDimension;
     /** Diverging route radius (straight turnouts). */
     divergingRadius?: PartDimension;
     /** Curved turnouts: the two concentric radii. */
@@ -1520,6 +1529,15 @@ declare const CODE55_RAIL_HEIGHT_INCHES = 0.055;
  *     #5     1¾″      4.75″     6.00″     3.00000″   1.25″       0.2500
  *     #7     ⁵⁄₈″     4⁷⁄₃₂″    6.00″     3.59375″   1.78125″    0.2545
  *     #10    ⁹⁄₁₆″    5.50″     8.00″     4.93750″   2.50″       0.2500
+ *     wye    points   frog V    overall   lead       past frog   diverging rail
+ *     #2.5   1⁵⁄₈″    4⅛″       6.50″     2.50000″   2.375″      2.4375″
+ *     #3.5   ¾″       3⁵⁄₃₂″    5.00″     2.40625″   1.84375″    1.9375″
+ *
+ * ⚠️ THE WYES ARE NOT ON THE ÷N TREND AND MUST NOT BE INTERPOLATED WITH THE
+ * TURNOUTS. lead ÷ N is 1.00 and 0.69 against the turnouts' flat 0.25, and the
+ * #2.5 is the LONGER part despite the sharper frog. Every size lookup here
+ * filters `kind === "turnout"`, which is what keeps them out — do not "fix"
+ * that filter to include wyes.
  *
  * Every one of these is Will Gage's, off a physical part, frog taken at the V.
  * ⚠️ The #7 lead SUPERSEDES Steve Branton's 3⅜″ (#173) — the library's founding
@@ -1883,6 +1901,8 @@ interface StoredTrackPart {
     frogOffsetSource?: string | null;
     overallLengthInches?: number | null;
     overallLengthSource?: string | null;
+    divergingLengthInches?: number | null;
+    divergingLengthSource?: string | null;
     leadInches?: number | null;
     leadSource?: string | null;
     outerRadiusInches?: number | null;
@@ -1916,6 +1936,17 @@ declare function storedPartToTrackPart(row: StoredTrackPart): TrackPart;
  *
  * The built-ins remain the floor: a part nobody has stored still resolves, so
  * geometry keeps working with no database at all.
+ *
+ * ⛔ THE TRAP THIS CREATES, and it has already been sprung once. Replacement is
+ * WHOLESALE, so a stored row that is merely INCOMPLETE deletes every dimension
+ * the built-in had. When both Atlas wyes were measured in 0.78.0 their stored
+ * rows still held null offsets from the original seed — so in production the
+ * new measurements were invisible and the 2056 still carried a `derived` lead
+ * the release had just retired. Nothing failed; the parts simply had no
+ * dimensions, which is indistinguishable from never having measured them.
+ *
+ * **Measuring a built-in is therefore not finished until the stored row is
+ * updated too.** Adding a dimension here and stopping ships nothing.
  */
 declare function mergeStoredParts(stored: StoredTrackPart[], library?: TrackPart[]): TrackPart[];
 /**

package/dist/index.js

@@ -1411,9 +1411,29 @@ var ATLAS_CODE55_N = [
     partNumbers: { single: "2056" },
     frogNumber: 2.5,
     lead: {
-      inches: 2.5 * TURNOUT_LEAD_INCHES_PER_FROG,
-      source: "derived",
-      note: "scaled from the measured #7 by frog number \u2014 not measured"
+      inches: 2.5,
+      source: "measured",
+      note: "Will Gage, physical Atlas 2056, 2026-07-26: points at 1\u2075\u2044\u2088\u2033, frog at 4\u215B\u2033 \u21D2 2.5\u2033 \u2014 the difference of the two offsets, per the rule above. \u26A0\uFE0F RETIRES a `derived` 1.205\u2033 (2.5 \xD7 TURNOUT_LEAD_INCHES_PER_FROG) \u2014 the real lead is more than DOUBLE it. Third independent refutation of the per-frog rule, and the largest: at low N it is not merely mis-sloped, it is nowhere near."
+    },
+    pointsOffset: {
+      inches: 1.625,
+      source: "measured",
+      note: "Will Gage, physical Atlas 2056 \u2014 end of tie strip to point tips (1\xB9\u2070\u2044\u2081\u2086\u2033)"
+    },
+    frogOffset: {
+      inches: 4.125,
+      source: "measured",
+      note: "Will Gage, physical Atlas 2056 \u2014 tie end to the apex of the V (4\xB2\u2044\u2081\u2086\u2033)"
+    },
+    overallLength: {
+      inches: 6.5,
+      source: "measured",
+      note: "Will Gage, physical Atlas 2056 \u2014 end to end"
+    },
+    divergingLength: {
+      inches: 2.4375,
+      source: "measured",
+      note: "Will Gage, physical Atlas 2056 \u2014 frog to the end of each diverging rail (2\u2077\u2044\u2081\u2086\u2033, both legs equal). \u2714 CROSS-CHECKS the frog: axial past-frog is 6.5 \u2212 4.125 = 2.375\u2033, and the rail is \xB9\u2044\u2081\u2086\u2033 longer than its own projection, which is the right sign and the right order of magnitude."
     }
   },
   {
@@ -1421,11 +1441,17 @@ var ATLAS_CODE55_N = [
     // wyes are different FROG NUMBERS, not a left/right pair of one part. The
     // 2056 half of that confirms the entry above; this is the one we didn't have.
     //
-    // No lead. The #2.5's is `derived` from a per-frog rule its own note records
-    // as refuted, so it rests on nothing — repeating that for a second part
-    // would be exactly how a number nobody checked becomes two. Without one,
-    // `leadInchesForSize` interpolates across the MEASURED parts at the wye's
-    // effective frog (3.5 × 2 = 7), which lands on the measured #7.
+    // Fully measured 2026-07-26, after one reading was caught and re-read.
+    //
+    // ⚠️ WORKED EXAMPLE OF WHY THIS LIBRARY CROSS-CHECKS. The frog first came in
+    // at 5⁵⁄₃₂″, which is PAST the part's own 5″ overall length — the apex would
+    // have sat beyond the end of the moulding. Two independent lines said ~3⅛″:
+    //   · points ¾″ + points→frog 2⅜″           = 3.125″
+    //   · overall 5″ − diverging rail 1³⁰⁄₃₂″    ≈ 3.06″ axial
+    // Queried, and Will re-read it as 3⁵⁄₃₂″ — a leading-digit slip, exactly what
+    // the arithmetic implied. That re-read is the number recorded here; it was
+    // NOT inferred. Had it been taken at face value the wye would have drawn its
+    // frog off the end of itself.
     id: "atlas-c55-n-wye-35",
     manufacturer: "Atlas",
     line: "Code 55",
@@ -1433,7 +1459,32 @@ var ATLAS_CODE55_N = [
     name: "#3.5 Wye",
     kind: "wye",
     partNumbers: { single: "2057" },
-    frogNumber: 3.5
+    frogNumber: 3.5,
+    lead: {
+      inches: 2.40625,
+      source: "measured",
+      note: "Will Gage, physical Atlas 2057, 2026-07-26: points at \xBE\u2033, frog at 3\u2075\u2044\u2083\u2082\u2033 \u21D2 2.40625\u2033 \u2014 the difference of the two offsets, per the rule above. His direct points\u2192frog reading was 2\u215C\u2033, \xB9\u2044\u2083\u2082\u2033 short; the offsets win. \u26A0\uFE0F RETIRES the reasoning in this entry's original note, which left the lead unset so `leadInchesForSize` would interpolate at the wye's effective frog (3.5 \xD7 2 = 7) and land on the measured #7's 3.59375\u2033. That substitution was 49% over."
+    },
+    pointsOffset: {
+      inches: 0.75,
+      source: "measured",
+      note: "Will Gage, physical Atlas 2057 \u2014 end of tie strip to point tips (\xBE\u2033)"
+    },
+    frogOffset: {
+      inches: 3.15625,
+      source: "measured",
+      note: "Will Gage, physical Atlas 2057 \u2014 tie end to the apex of the V (3\u2075\u2044\u2083\u2082\u2033). This is his RE-READ after the first reading (5\u2075\u2044\u2083\u2082\u2033) was queried for landing past the part's 5\u2033 end; see the comment above this entry."
+    },
+    overallLength: {
+      inches: 5,
+      source: "measured",
+      note: "Will Gage, physical Atlas 2057 \u2014 end to end"
+    },
+    divergingLength: {
+      inches: 1.9375,
+      source: "measured",
+      note: "Will Gage, physical Atlas 2057 \u2014 frog to the end of each diverging rail (1\xB3\u2070\u2044\u2083\u2082\u2033, both legs equal). \u2714 THE READING THAT CAUGHT THE BAD FROG: against the first 5\u2075\u2044\u2083\u2082\u2033 the axial past-frog is NEGATIVE, which a 1.9375\u2033 rail flatly contradicts. Against the re-read 3\u2075\u2044\u2083\u2082\u2033 it is 5 \u2212 3.15625 = 1.84375\u2033, so the rail runs \xB3\u2044\u2083\u2082\u2033 longer than its own projection \u2014 right sign, same order as the 2056's \xB9\u2044\u2081\u2086\u2033. (The two differ because this is a cos() residual on a small angle, which is why the field doc says not to read an angle back out of it.)"
+    }
   },
   {
     id: "atlas-c55-n-curved-21-15",
@@ -1789,6 +1840,8 @@ function storedPartToTrackPart(row) {
   if (frog) part.frogOffset = frog;
   const overall = dim(row.overallLengthInches, row.overallLengthSource);
   if (overall) part.overallLength = overall;
+  const diverging = dim(row.divergingLengthInches, row.divergingLengthSource);
+  if (diverging) part.divergingLength = diverging;
   if (lead) part.lead = lead;
   const outer = dim(row.outerRadiusInches, row.radiusSource);
   const inner = dim(row.innerRadiusInches, row.radiusSource);