Fix writer examples

desktop_env/evaluators/metrics/docs.py

@@ -1,13 +1,16 @@
 import logging
 import os
-import xml.etree.ElementTree as ET
-import zipfile
 import re
+import xml.etree.ElementTree as ET
 from typing import List, Dict, Any
 
 from docx import Document
 from docx.enum.text import WD_PARAGRAPH_ALIGNMENT
 from docx.shared import RGBColor
+from odf.opendocument import load
+from odf.text import P
+from odf.text import Span
+import zipfile
 
 logger = logging.getLogger("desktopenv.metric.docs")
 
@@ -51,12 +54,39 @@ def contains_page_break(docx_file):
 
 
 def compare_docx_files(file1, file2, ignore_blanks=True):
-    doc1 = Document(file1)
-    doc2 = Document(file2)
+    def get_paragraph_texts_odt(document):
+        paragraphs = document.getElementsByType(P)
+        paragraph_texts = []
+        for paragraph in paragraphs:
+            text_parts = []
+            for node in paragraph.childNodes:
+                if node.nodeType == node.TEXT_NODE:
+                    text_parts.append(node.data)
+                elif node.nodeType == node.ELEMENT_NODE and node.tagName == 'text:span':
+                    # Assuming direct text content in <text:span>, for simplicity
+                    for child in node.childNodes:
+                        if child.nodeType == child.TEXT_NODE:
+                            text_parts.append(child.data)
+            paragraph_texts.append(''.join(text_parts))
+        return paragraph_texts
 
-    doc1_paragraphs = [p.text for p in doc1.paragraphs]
-    doc2_paragraphs = [p.text for p in doc2.paragraphs]
+    # Determine file types and load documents
+    if file1.endswith('.docx') and file2.endswith('.docx'):
+        doc1 = Document(file1)
+        doc2 = Document(file2)
+        doc1_paragraphs = [p.text for p in doc1.paragraphs]
+        doc2_paragraphs = [p.text for p in doc2.paragraphs]
+    elif file1.endswith('.odt') and file2.endswith('.odt'):
+        doc1 = load(file1)
+        doc2 = load(file2)
+        doc1_paragraphs = get_paragraph_texts_odt(doc1)
+        doc2_paragraphs = get_paragraph_texts_odt(doc2)
+    else:
+        # Unsupported file types or mismatch
+        print("Unsupported file types or mismatch between file types.")
+        return 0
 
+    # Process and compare documents
     if ignore_blanks:
         text1 = re.sub(r'\s+', ' ', '\n'.join(doc1_paragraphs)).strip()
         text2 = re.sub(r'\s+', ' ', '\n'.join(doc2_paragraphs)).strip()
@@ -64,19 +94,16 @@ def compare_docx_files(file1, file2, ignore_blanks=True):
             return 0
     else:
         if len(doc1_paragraphs) != len(doc2_paragraphs):
-            # print(len(doc1_paragraphs))
-            # print(len(doc2_paragraphs))
             return 0
 
         # Compare each paragraph
         for p1, p2 in zip(doc1_paragraphs, doc2_paragraphs):
             if p1 != p2:
-                # print(p1)
-                # print(p2)
                 return 0
 
     return 1
 
+
 def compare_init_lines(file1, file2):
     doc1 = Document(file1)
     doc2 = Document(file2)
@@ -93,6 +120,7 @@ def compare_init_lines(file1, file2):
 
     return 1
 
+
 def compare_docx_tables(docx_file1, docx_file2):
     doc1 = Document(docx_file1)
     doc2 = Document(docx_file2)
@@ -217,6 +245,7 @@ def compare_contains_image(docx_file1, docx_file2):
                 return 0
     return 1
 
+
 # file1 = 'path/to/file1.docx'
 # file2 = 'path/to/file2.docx'
 
@@ -253,26 +282,25 @@ def evaluate_colored_words_in_tables(file_path1, file_path2):
 
 
 def check_highlighted_words(file_path1, file_path2):
-    # if not compare_docx_files(file_path1, file_path2):
-    #     return 0
-        
-    # Extract content.xml from the .odt file
-    extract_dir = file_path1 + "_extracted"
-    with zipfile.ZipFile(file_path1, 'r') as zip_ref:
-        zip_ref.extractall(extract_dir)
-    content_xml_path = os.path.join(extract_dir, 'content.xml')
-    with open(content_xml_path, 'r', encoding="utf-8") as file:
-        content_xml = file.read()
-
-    # Check for yellow highlights in the content.xml
-    yellow_highlight_pattern = re.compile(r'(.{0,50}background-color="#ffff00"[^>]*>.{0,50})')
-    yellow_highlight_matches = yellow_highlight_pattern.findall(content_xml)
-
-    # Return True if yellow highlights are NOT found, otherwise True
-    if yellow_highlight_matches:
+    if not compare_docx_files(file_path1, file_path2):
         return 0
-    else:
-        return 1
+
+    doc = load(file_path1)
+    highlighted = False
+
+    for span in doc.getElementsByType(Span):
+        style_name = span.getAttribute('stylename')
+        if style_name:
+            for automatic_style in doc.automaticstyles.childNodes:
+                if automatic_style.getAttribute('name') == style_name:
+                    for property in automatic_style.childNodes:
+                        if property.getAttribute('backgroundcolor') == '#ffff00':
+                            highlighted = True
+                            break
+            if highlighted:
+                break
+
+    return 0 if highlighted else 1
 
 
 def evaluate_strike_through_last_paragraph(file_path1, file_path2):
@@ -318,9 +346,9 @@ def evaluate_spacing(file_path):
     body_spacing = document.paragraphs[1].paragraph_format.line_spacing
     conclusion_spacing = document.paragraphs[2].paragraph_format.line_spacing
     if (introduction_spacing == 1.0 and body_spacing == 2.0 and conclusion_spacing == 1.5):
-      return 1
+        return 1
     else:
-      return 0
+        return 0
 
 
 def check_italic_font_size_14(path1, path2):
@@ -358,13 +386,14 @@ def evaluate_alignment(docx_path):
             second_part = ' '.join(words[3:])
 
             # Check if the sentence structure matches the pattern: first part + large space/tab + second part
-            if not (first_part in sentence and second_part in sentence and sentence.find(first_part) < sentence.find(second_part)):
+            if not (first_part in sentence and second_part in sentence and sentence.find(first_part) < sentence.find(
+                    second_part)):
                 return 0  # The sentence does not meet the alignment criteria
 
     return 1  # All sentences meet the alignment criteria
 
 
-def get_unique_train_ids(initial_file): #fixed standard
+def get_unique_train_ids(initial_file):  # fixed standard
     doc = Document(initial_file)
     train_ids = set()
     processed_lines = 0
@@ -374,11 +403,12 @@ def get_unique_train_ids(initial_file): #fixed standard
         if len(line_parts) == 4:
             train_id = line_parts[1].strip()
             if train_id not in train_ids:
-              train_ids.add(train_id)
-              processed_lines += 1
+                train_ids.add(train_id)
+                processed_lines += 1
 
     return train_ids, processed_lines
 
+
 def check_no_duplicates(initial_file, processed_file):
     # Open the document
     train_ids_ini, ini_lines = get_unique_train_ids(initial_file)
@@ -405,6 +435,7 @@ def check_no_duplicates(initial_file, processed_file):
     # No duplicates found and at least one valid line was processed
     return 1
 
+
 def compare_docx_lines(file1, file2):
     # Read the text of the document, line by line
     doc1 = Document(file1)
@@ -416,21 +447,46 @@ def compare_docx_lines(file1, file2):
     # print(doc2_lines)
 
     # Convert the list of lines to sets and compare
-    return set(doc1_lines) == set(doc2_lines)
+    if set(doc1_lines) == set(doc2_lines):
+        return 1
+    else:
+        return 0
 
+
+# Docx file saved in the ubuntu cannot use this function to compare highlight, don't know why, deprecated
 def compare_highlighted_text(file1, file2):
-    def extract_highlighted_text(doc):
-        highlighted_text = []
-        # Iterate through each run in each paragraph to check for highlight
-        for paragraph in doc.paragraphs:
-            for run in paragraph.runs:
-                if run.font.highlight_color:  # Checks if the run is highlighted
-                    highlighted_text.append(run.text.strip())
-        return highlighted_text
+    def extract_highlighted_text(file_path):
+        highlighted_texts = []
+
+        # Open the .docx file as a zip file and read the document.xml
+        with zipfile.ZipFile(file_path, 'r') as docx:
+            with docx.open('word/document.xml') as document_xml:
+                tree = ET.parse(document_xml)
+                root = tree.getroot()
+
+        # Define the namespaces
+        namespaces = {
+            'w': 'http://schemas.openxmlformats.org/wordprocessingml/2006/main',
+        }
+
+        # Find all runs with highlight property
+        for run in root.findall('.//w:r', namespaces):
+            highlight = run.find('.//w:highlight', namespaces)
+            if highlight is not None and highlight.get(
+                    '{http://schemas.openxmlformats.org/wordprocessingml/2006/main}val') != 'none':
+                text = run.find('.//w:t', namespaces)
+                if text is not None:
+                    highlighted_texts.append(text.text)
+
+        return highlighted_texts
 
     # Read the highlighted text from both documents
-    doc1_highlighted = extract_highlighted_text(Document(file1))
-    doc2_highlighted = extract_highlighted_text(Document(file2))
+    doc1_highlighted = extract_highlighted_text(file1)
+    doc2_highlighted = extract_highlighted_text(file2)
 
     # Compare the sets of highlighted text to check if they are the same
-    return set(doc1_highlighted) == set(doc2_highlighted)
+    if set(doc1_highlighted) == set(doc2_highlighted):
+        return 1
+    else:
+        return 0
+

evaluation_examples/examples/libreoffice_writer/41c621f7-3544-49e1-af8d-dafd0f834f75.json

@@ -1,7 +1,7 @@
 {
   "id": "41c621f7-3544-49e1-af8d-dafd0f834f75",
   "snapshot": "libreoffice_writer",
-  "instruction": "I am adding comments which begin with \"#\" right beside sentences my students have written. I want to make my comments to stand out more, so I am highlighting my comments to yellow. Could you help me on this? It is hard for me to color comments one by one.",
+  "instruction": "I added annotations starting with a \"#\" next to the sentences that my students wrote. I want my annotations to stand out more, so I highlighted them in yellow. Can you help me with this favor? I'm having a hard time coloring the annotations one by one. By the way, remember to remove the # sign after highlighting the text part. Thanks!",
   "source": "https://superuser.com/questions/1668018/how-to-auto-format-lines-in-libre-office-writer",
   "config": [
     {
@@ -52,10 +52,7 @@
         }
       }
     ],
-    "func": [
-      "compare_highlighted_text",
-      "compare_docx_files"
-    ],
+    "func": "compare_highlighted_text",
     "expected": {
       "type": "cloud_file",
       "path": "https://drive.usercontent.google.com/download?id=1x2gplRDN1AOnRWiqfqmJqN9JTQ_x4sPu&export=download&authuser=0&confirm=t&uuid=ea6b0a61-fd06-4823-b253-05473c77c192&at=APZUnTUxWWFB_TOKLR9chabErm7b:1706018376493",

evaluation_examples/examples/libreoffice_writer/6a33f9b9-0a56-4844-9c3f-96ec3ffb3ba2.json

@@ -9,7 +9,7 @@
       "parameters": {
         "files": [
           {
-            "url": "https://drive.google.com/uc?id=1Ul4mtQ4SpUNLVDlaiJThPprBe6WTGZ2L&export=download",
+            "url": "https://drive.usercontent.google.com/download?id=1_1uItoxJsp8PgBqaED0U3bOfKxIweNkJ&export=download&authuser=0&confirm=t&uuid=facec3ed-2fd6-49f7-aeda-95feaa38f45c&at=APZUnTU6B0xoLG-kI0V2WZ74oSI6:1706544439742",
             "path": "Desktop/sample-recruitment-phone-script.odt"
           }
         ]
@@ -55,7 +55,7 @@
     "func": "check_highlighted_words",
     "expected": {
       "type": "cloud_file",
-      "path": "https://drive.google.com/uc?id=12iMkgCYuUyhKUXux96kANLIeud0Wz9ct&export=download",
+      "path": "https://drive.usercontent.google.com/download?id=1Qxz3EjHuiLa8yZaCr1DJHwEOSd6U-cjY&export=download&authuser=0&confirm=t&uuid=c502680e-51f9-4e4f-90d0-58b123a09b3d&at=APZUnTXN2UX9EuvB1PZoJ5koJwmQ:1706544440969",
       "dest": "sample-recruitment-phone-script_Gold.odt"
     },
     "result": {

requirements.txt

@@ -36,5 +36,5 @@ backoff
 formulas
 pydrive
 fastdtw
-
+odfpy
 openai