@jbrowse/plugin-gff3 5.0.0-beta.1 → 5.0.0-beta.2

package/esm/Gff3Adapter/Gff3Adapter.js

@@ -22,17 +22,16 @@ export default class Gff3Adapter extends BaseFeatureDataAdapter {
             // that resident set small (8.5x on GENCODE-shaped input), and the
             // render path reads only a handful of attributes anyway — see
             // Gff3Feature.
-            (lines, refName) => {
-                const features = parseLinesLazy(lines);
-                // stamped in place rather than through `{...feature, uniqueId}`:
-                // these are freshly parsed objects nobody else holds, and the spread
-                // copied every attribute of every top-level feature in the file to
-                // add one key
-                for (let i = 0; i < features.length; i++) {
-                    features[i].uniqueId = `${this.id}-${refName}-${i}`;
-                }
-                return features;
-            }, 'Parsing GFF data');
+            //
+            // The id the tree entry carries is minted here, beside the feature
+            // rather than stamped onto it, so the parsed feature is the library's
+            // shape and Gff3Feature serializes it identically for both adapters.
+            (lines, refName) => parseLinesLazy(lines).map((feature, i) => ({
+                start: feature.start,
+                end: feature.end,
+                feature,
+                uniqueId: `${this.id}-${refName}-${i}`,
+            })), 'Parsing GFF data');
             return { header: headerLines.join('\n'), intervalTreeMap };
         },
     });
@@ -52,8 +51,11 @@ export default class Gff3Adapter extends BaseFeatureDataAdapter {
             const { intervalTreeMap } = await this.loadData(opts);
             const tree = intervalTreeMap[refName];
             if (tree) {
-                for (const f of tree(opts.statusCallback).search([start, end])) {
-                    observer.next(new Gff3Feature(f, f.uniqueId));
+                for (const { feature, uniqueId } of tree(opts.statusCallback).search([
+                    start,
+                    end,
+                ])) {
+                    observer.next(new Gff3Feature(feature, uniqueId));
                 }
             }
             observer.complete();

package/esm/Gff3Feature.d.ts

@@ -1,9 +1,5 @@
 import type { Feature, SimpleFeatureSerialized } from '@jbrowse/core/util/simpleFeature';
 import type { LazyGffFeature } from 'gff-nostream';
-/** A parsed GFF3 feature carrying the id the adapter minted for it. */
-export type IdentifiedGffFeature = LazyGffFeature & {
-    uniqueId: string;
-};
 /**
  * A {@link Feature} over a GFF3 feature whose attributes are still raw column-9
  * text, resolving each one only when something asks for it.
@@ -26,7 +22,7 @@ export declare class Gff3Feature implements Feature {
     private uniqueId;
     private parentFeature?;
     private inflated?;
-    constructor(data: IdentifiedGffFeature | LazyGffFeature, uniqueId: string, parentFeature?: Feature | undefined);
+    constructor(data: LazyGffFeature, uniqueId: string, parentFeature?: Feature | undefined);
     /**
      * A switch rather than a lookup table or a `data[name]` fallthrough: the
      * fixed columns are read tens of times per feature by the layout pass, and

package/esm/Gff3TabixAdapter/Gff3TabixAdapter.js

@@ -7,7 +7,7 @@ import { openLocation, openTabixIndexFilehandle } from '@jbrowse/core/util/io';
 import { doesIntersect2 } from '@jbrowse/core/util/range';
 import { ObservableCreate } from '@jbrowse/core/util/rxjs';
 import { readTabixLinesRedispatched } from '@jbrowse/core/util/tabix';
-import { parseRecordsLazy } from 'gff-nostream';
+import { hasIdAttribute, parseRecordsLazy } from 'gff-nostream';
 import { Gff3Feature } from '../Gff3Feature.js';
 export default class Gff3TabixAdapter extends BaseFeatureDataAdapter {
     configure = cachedSetup({
@@ -47,7 +47,26 @@ export default class Gff3TabixAdapter extends BaseFeatureDataAdapter {
         return ObservableCreate(async (observer) => {
             try {
                 const { gff, dontRedispatchSet } = await this.configure(opts);
-                const lines = await readTabixLinesRedispatched(gff, query, dontRedispatchSet, opts);
+                // The bound exists to reach a record's children, so what disqualifies a
+                // record is having none. `ID` answers that exactly — nothing can name a
+                // parent that has no name — and gff-nostream decides it with the scan
+                // its linker uses, so a record the tree would attach children to is
+                // never one the bound leaves out. `dontRedispatch` can only approximate
+                // the same question by type: every NCBI `GCF_*_genomic.gff.gz` opens
+                // each reference with a chromosome-long `match` record no list named,
+                // and none of them has an `ID`. The list covers what `ID` cannot see: a
+                // record carrying an `ID` that nothing references, which hosted hg19
+                // RefSeq's chromosome-long `region` is.
+                const canHaveChildren = (line) => hasIdAttribute(line.line) && !dontRedispatchSet.has(line.type);
+                const lines = await readTabixLinesRedispatched(gff, query, 
+                // A caller reading only top-level features is owed nothing by the
+                // flanks and pays for them anyway: they complete SUBFEATURE lists, and
+                // a top-level feature overlapping the query is in the query's own read
+                // already, since tabix returns every overlapping line and a child is
+                // contained in its parent. On an NCBI `GCF_*_genomic.gff.gz`, whose
+                // chromosome-long `match` widens the bound at every window, that is
+                // 193,008 lines parsed to keep 3 features — 2734 ms against 8 ms.
+                opts.topLevelOnly ? () => false : canHaveChildren, opts);
                 // emit only top-level features intersecting the original query. the
                 // byte offset stays on our own record and is used purely to mint a
                 // stable id, so it never pollutes the feature's data

package/esm/Gff3TabixAdapter/configSchema.d.ts

@@ -18,6 +18,11 @@ export declare function normalizeSnapshot(snap: Record<string, unknown>): Record
  * ```
  */
 declare const Gff3TabixAdapter: import("@jbrowse/core/configuration/configurationSchema").ConfigurationSchemaType<{
+    readonly densityAdapter: {
+        readonly type: 'frozen';
+        readonly defaultValue: null;
+        readonly description: 'optional quantitative sub-adapter (e.g. a BigWigAdapter over a features-per-bin bigWig) drawn as a density band where the region is too large to fetch features; null disables it';
+    };
     /**
      * #slot
      * location of the bgzip-compressed GFF3, sorted by position. Must be bgzip

package/esm/Gff3TabixAdapter/configSchema.js

@@ -1,4 +1,5 @@
 import { ConfigurationSchema, expandTabixShorthand, tabixIndexFields, } from '@jbrowse/core/configuration';
+import { densityAdapterConfigSchemaFields } from '@jbrowse/core/data_adapters/BaseAdapter';
 export function normalizeSnapshot(snap) {
     return expandTabixShorthand(snap, 'gffGzLocation');
 }
@@ -56,6 +57,7 @@ const Gff3TabixAdapter = ConfigurationSchema('Gff3TabixAdapter', {
             'scaffold',
         ],
     },
+    ...densityAdapterConfigSchemaFields,
 }, {
     explicitlyTyped: true,
     /**

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jbrowse/plugin-gff3",
-  "version": "5.0.0-beta.1",
+  "version": "5.0.0-beta.2",
   "type": "module",
   "description": "JBrowse 2 gff3.",
   "keywords": [
@@ -23,10 +23,10 @@
   "dependencies": {
     "@gmod/tabix": "^3.8.2",
     "@jbrowse/mobx-state-tree": "^6.5.1",
-    "gff-nostream": "^5.3.0",
+    "gff-nostream": "^5.4.0",
     "mobx": "^7.0.3",
     "rxjs": "^7.8.2",
-    "@jbrowse/core": "^5.0.0-beta.1"
+    "@jbrowse/core": "^5.0.0-beta.2"
   },
   "publishConfig": {
     "access": "public"