Merge pull request #775 from JaredReisinger/fix-doc-offset-start

Fix doc/result offset start
spencermountain · Aug 10, 2020 · 2b90ad5 · 2b90ad5
2 parents 3fffa74 + 1df7d92
commit 2b90ad5
Show file tree

Hide file tree

Showing 2 changed files with 29 additions and 30 deletions.
diff --git a/src/Doc/methods/output/_offset.js b/src/Doc/methods/output/_offset.js
@@ -40,15 +40,16 @@ const calcOffset = function(doc, result, options) {
       //   return n
       // }, 0)
 
-      // offset information for the entire doc starts at the first term, and
-      // is as long as the whole text (note that there may be an issue where
-      // leading punctuation is counted in the doc text length, but is
-      // *excluded* from the term[0] start position)
-      o.offset = Object.assign(
-        {},
-        o.terms[0].offset,
-        { length: o.text.length }
-      )
+      // The offset information for the entire doc starts at (or just before)
+      // the first term, and is as long as the whole text.  The code originally
+      // copied the entire offset value from terms[0], but since we're now
+      // overriding 2 of the three fields, it's cleaner to just create an all-
+      // new object and not pretend it's "just" the same as terms[0].
+      o.offset = {
+        index: o.terms[0].offset.index,
+        start: o.terms[0].offset.start - o.text.indexOf(o.terms[0].text),
+        length: o.text.length
+      }
     })
   }
 }

diff --git a/tests/output/offset.test.js b/tests/output/offset.test.js
@@ -26,7 +26,7 @@ test('offset-punctuation', function(t) {
   let doc = nlp(`one (two two) more `)
   let m = doc.match('two two')
   let obj = m.json({ offset: true, terms: false })[0]
-  t.equal(obj.offset.start, 5, '4 two-start')
+  t.equal(obj.offset.start, 4, '4 two-start')
   t.equal(obj.offset.length, 9, '4 two-length')
 
   doc = nlp(`0123, 678`)
@@ -43,14 +43,14 @@ test('offset-terms', function(t) {
   let doc = nlp(`hello world`)
   let obj = doc.json({ offset: true, terms: true })[0]
 
-  t.equal(obj.offset.start, 0, 'doc-start')
-  t.equal(obj.offset.length, 11, 'doc-length')
+  t.equal(obj.offset.start, 0, '6 doc-start')
+  t.equal(obj.offset.length, 11, '6 doc-length')
 
-  t.equal(obj.terms[0].offset.start, 0, 'term 0-start')
-  t.equal(obj.terms[0].offset.length, 5, 'term 0-length')
+  t.equal(obj.terms[0].offset.start, 0, '6 term 0-start')
+  t.equal(obj.terms[0].offset.length, 5, '6 term 0-length')
 
-  t.equal(obj.terms[1].offset.start, 6, 'term 0-start')
-  t.equal(obj.terms[1].offset.length, 5, 'term 0-length')
+  t.equal(obj.terms[1].offset.start, 6, '6 term 1-start')
+  t.equal(obj.terms[1].offset.length, 5, '6 term 1-length')
 
   t.end()
 })
@@ -59,14 +59,14 @@ test('offset-terms-whitespace', function(t) {
   let doc = nlp(` hello world`)
   let obj = doc.json({ offset: true, terms: true })[0]
 
-  t.equal(obj.offset.start, 1, 'doc-start')
-  t.equal(obj.offset.length, 11, 'doc-length')
+  t.equal(obj.offset.start, 1, '7 doc-start')
+  t.equal(obj.offset.length, 11, '7 doc-length')
 
-  t.equal(obj.terms[0].offset.start, 1, 'term 0-start')
-  t.equal(obj.terms[0].offset.length, 5, 'term 0-length')
+  t.equal(obj.terms[0].offset.start, 1, '7 term 0-start')
+  t.equal(obj.terms[0].offset.length, 5, '7 term 0-length')
 
-  t.equal(obj.terms[1].offset.start, 7, 'term 0-start')
-  t.equal(obj.terms[1].offset.length, 5, 'term 0-length')
+  t.equal(obj.terms[1].offset.start, 7, '7 term 1-start')
+  t.equal(obj.terms[1].offset.length, 5, '7 term 1-length')
 
   t.end()
 })
@@ -75,16 +75,14 @@ test('offset-terms-punctuation', function(t) {
   let doc = nlp(`"hello world`)
   let obj = doc.json({ offset: true, terms: true })[0]
 
-  // The doc-level offset should perhaps be 0->12 or 1->11... but 1->12 is not
-  // really sane.  This test will need to change if/when that gets figured out.
-  t.equal(obj.offset.start, 1, 'doc-start')    // <==== arguably wrong!
-  t.equal(obj.offset.length, 12, 'doc-length') // <==== arguably wrong!
+  t.equal(obj.offset.start, 0, '8 doc-start')
+  t.equal(obj.offset.length, 12, '8 doc-length')
 
-  t.equal(obj.terms[0].offset.start, 1, 'term 0-start')
-  t.equal(obj.terms[0].offset.length, 5, 'term 0-length')
+  t.equal(obj.terms[0].offset.start, 1, '8 term 0-start')
+  t.equal(obj.terms[0].offset.length, 5, '8 term 0-length')
 
-  t.equal(obj.terms[1].offset.start, 7, 'term 0-start')
-  t.equal(obj.terms[1].offset.length, 5, 'term 0-length')
+  t.equal(obj.terms[1].offset.start, 7, '8 term 1-start')
+  t.equal(obj.terms[1].offset.length, 5, '8 term 1-length')
 
   t.end()
 })