tag optimisation and #352 fix

scratch.js

@@ -19,17 +19,23 @@ const fresh = require('./test/unit/lib/freshPrince.js');
 // r.contractions().debug();
 
 //===timer
-console.time('parse');
-let r = nlp(fresh);
-console.timeEnd('parse');
-
-console.time('match');
-r.match('#Determiner (story|thing|#Adjective)', true);
-console.timeEnd('match');
-
-console.time('tag');
-r.tag('#Person');
-console.timeEnd('tag');
-
-// r = nlp('6 am').debug();
-// console.log(r.people().data());
+// console.time('parse');
+// let r = nlp(fresh);
+// console.timeEnd('parse');
+//
+// console.time('match');
+// r.match('#Determiner (story|thing|#Adjective)', true);
+// console.timeEnd('match');
+//
+// console.time('tag');
+// r.tag('#Person');
+// console.timeEnd('tag');
+
+let r = nlp('i\'d contemplate');
+console.log(r.normalize().out('text'));
+
+r = nlp('i\'d contemplated');
+console.log(r.normalize().out('text'));
+
+r = nlp('i\'d really contemplated');
+console.log(r.normalize().out('text'));

src/result/methods/loops.js

@@ -21,7 +21,7 @@ const methods = [
 
   'tagger',
 
-  'tag',
+  // 'tag',
   'unTag',
 ];
 
@@ -34,6 +34,18 @@ const addMethods = (Text) => {
       return this;
     };
   });
+
+  //add an extra optimisation for tag method
+  Text.prototype.tag = function() {
+    //fail-fast optimisation
+    if (this.list.length === 0) {
+      return this;
+    }
+    for(let i = 0; i < this.list.length; i++) {
+      this.list[i].tag.apply(this.list[i], arguments);
+    }
+    return this;
+  };
 };
 
 module.exports = addMethods;

src/result/methods/match/index.js

@@ -51,7 +51,7 @@ const splitMethods = (Text) => {
     /** do a regex-like search through terms and return a subset */
     match: function (reg, verbose) {
       //fail-fast
-      if (reg === undefined || reg === null) {
+      if (this.list.length === 0 || reg === undefined || reg === null) {
         let parent = this.parent || this;
         return new Text([], this.lexicon, parent);
       }

src/tagger/contraction/03-easyOnes.js

@@ -5,7 +5,7 @@ const split = require('./split');
 //the formulaic contraction types:
 const easy_ends = {
   'll': 'will',
-  'd': 'would',
+  // 'd': 'would',
   've': 'have',
   're': 'are',
   'm': 'am',
@@ -16,6 +16,7 @@ const easy_ends = {
 };
 
 
+
 //unambiguous contractions, like "'ll"
 const easyOnes = (ts) => {
   for(let i = 0; i < ts.terms.length; i++) {
@@ -25,6 +26,26 @@ const easyOnes = (ts) => {
     }
     let parts = split(ts.terms[i]);
     if (parts) {
+
+      //handle i'd -> 'i would' vs 'i had'
+      if (parts.end = 'd') {
+        //assume 'would'
+        let arr = [
+          parts.start,
+          'would'
+        ];
+        //if next verb is past-tense, choose 'had'
+        if (ts.terms[i + 1] && ts.terms[i + 1].tags.PastTense) {
+          arr[1] = 'had';
+        }
+        //also support '#Adverb #PastTense'
+        if (ts.terms[i + 2] && ts.terms[i + 2].tags.PastTense && ts.terms[i + 1].tags.Adverb) {
+          arr[1] = 'had';
+        }
+        ts = fixContraction(ts, arr, i);
+        i += 1;
+      }
+
       //make sure its an easy one
       if (easy_ends[parts.end]) {
         let arr = [

src/terms/match/index.js

@@ -10,7 +10,11 @@ const matchMethods = (Terms) => {
 
     //support regex-like whitelist-match
     match: function (reg, verbose) {
-      //fail-fast
+      //fail-fast #1
+      if (this.terms.length === 0) {
+        return new Text([], this.lexicon, this.parent);
+      }
+      //fail-fast #2
       if (!reg) {
         return new Text([], this.lexicon, this.parent);
       }
@@ -23,6 +27,10 @@ const matchMethods = (Terms) => {
 
     /**return first match */
     matchOne: function (str) {
+      //fail-fast
+      if (this.terms.length === 0) {
+        return null;
+      }
       let regs = syntax(str);
       for (let t = 0; t < this.terms.length; t++) {
         //don't loop through if '^'

src/terms/match/lib/fastPass.js

@@ -1,7 +1,7 @@
 'use strict';
 //
 //find easy reasons to skip running the full match on this
-const fastPass = (ts, regs, verbose) => {
+const fastPass = (ts, regs) => {
   for(let i = 0; i < regs.length; i++) {
     let reg = regs[i];
     let found = false;
@@ -10,9 +10,6 @@ const fastPass = (ts, regs, verbose) => {
     }
     //look-for missing term-matches
     if (reg.normal !== undefined) {
-      if (verbose) {
-        console.log(reg);
-      }
       for(let o = 0; o < ts.terms.length; o++) {
         if (ts.terms[o].normal === reg.normal || ts.terms[o].silent_term === reg.normal) {
           found = true;

test/unit/subset/contractions/basic.test.js

@@ -91,3 +91,14 @@ test('contracted', function(t) {
   t.equal(str, `I will`, 'expanded');
   t.end();
 });
+
+test('would-or-did', function(t) {
+  var r = nlp(`i'd contemplate`);
+  var str = r.contractions().expand().all().out('text');
+  t.equal(str, `i would contemplate`, 'i-would');
+
+  r = nlp(`i'd contemplated`);
+  str = r.contractions().expand().all().out('text');
+  t.equal(str, `i had contemplated`, 'i-had');
+  t.end();
+});