Add cleaner way to load Data.from_line_format without TagReader, allo…

…wing to remove unused data processing code in the future

aste/data_utils.py

@@ -1,3 +1,4 @@
+import ast
 import copy
 import json
 import os
@@ -197,7 +198,27 @@ def as_text(self) -> str:
             tokens[t.t_end] = tokens[t.t_end] + "]"
         return " ".join(tokens)
 
-    def as_line_format(self) -> str:
+    @classmethod
+    def from_line_format(cls, text: str):
+        front, back = text.split("#### #### ####")
+        tokens = front.split(" ")
+        triples = []
+
+        for a, b, label in ast.literal_eval(back):
+            t = SentimentTriple(
+                t_start=a[0],
+                t_end=a[0] if len(a) == 1 else a[-1],
+                o_start=b[0],
+                o_end=b[0] if len(b) == 1 else b[-1],
+                label=label,
+            )
+            triples.append(t)
+
+        return cls(
+            tokens=tokens, triples=triples, id=0, pos=[], weight=1, is_labeled=True
+        )
+
+    def to_line_format(self) -> str:
         # ([1], [4], 'POS')
         # ([1,2], [4], 'POS')
         triplets = []
@@ -210,7 +231,11 @@ def as_line_format(self) -> str:
                     parts.append([start, end])
             parts.append(f"{t.label}")
             triplets.append(tuple(parts))
-        return " ".join(self.tokens) + "#### #### ####" + str(triplets) + "\n"
+
+        line = " ".join(self.tokens) + "#### #### ####" + str(triplets) + "\n"
+        assert self.from_line_format(line).tokens == self.tokens
+        assert self.from_line_format(line).triples == self.triples
+        return line
 
 
 class Data(BaseModel):
@@ -227,13 +252,9 @@ def load(self):
             path = self.root / f"{self.data_split}.txt"
             if self.full_path:
                 path = self.full_path
-            instances = TagReader.read_inst(
-                file=path,
-                is_labeled=self.is_labeled,
-                number=self.num_instances,
-                opinion_offset=self.opinion_offset,
-            )
-            self.sentences = [Sentence.from_instance(x) for x in instances]
+
+            with open(path) as f:
+                self.sentences = [Sentence.from_line_format(line) for line in f]
 
     @classmethod
     def load_from_full_path(cls, path: str):
@@ -246,7 +267,7 @@ def save_to_path(self, path: str):
         Path(path).parent.mkdir(exist_ok=True, parents=True)
         with open(path, "w") as f:
             for s in self.sentences:
-                f.write(s.as_line_format())
+                f.write(s.to_line_format())
 
         data = Data.load_from_full_path(path)
         assert data.sentences is not None
@@ -432,7 +453,32 @@ def analyze(self):
         print("#" * 80)
 
 
+def test_from_line_format(path: str = "aste/data/triplet_data/14lap/train.txt"):
+    print("\nCompare old TagReader with new Sentence.from_line_format")
+    instances = TagReader.read_inst(
+        file=path,
+        is_labeled=False,
+        number=-1,
+        opinion_offset=3,
+    )
+    a = Data(
+        root=Path(),
+        data_split=SplitEnum.test,
+        sentences=[Sentence.from_instance(x) for x in instances],
+    )
+
+    assert a.sentences is not None
+    with open(path) as f:
+        for i, line in enumerate(f):
+            s = Sentence.from_line_format(line)
+            assert s.tokens == a.sentences[i].tokens
+            set_a = set(t.json() for t in a.sentences[i].triples)
+            set_b = set(t.json() for t in s.triples)
+            assert set_a == set_b
+
+
 def test_save_to_path(path: str = "aste/data/triplet_data/14lap/train.txt"):
+    print("\nEnsure that Data.save_to_path works properly")
     path_temp = "temp.txt"
     data = Data.load_from_full_path(path)
     data.save_to_path(path_temp)