|
| 1 | +"""Test DocTags serialization filtering functionality.""" |
| 2 | + |
| 3 | +from docling_core.transforms.serializer.doctags import ( |
| 4 | + DocTagsDocSerializer, |
| 5 | + DocTagsParams, |
| 6 | + create_task_filtered_params, |
| 7 | +) |
| 8 | + |
| 9 | + |
| 10 | +def test_create_task_filtered_params_defaults(): |
| 11 | + """Test default behavior.""" |
| 12 | + params = create_task_filtered_params([]) |
| 13 | + assert params.include_ocr is True |
| 14 | + assert params.include_layout is True |
| 15 | + assert params.include_otsl is True |
| 16 | + assert params.include_code is True |
| 17 | + assert params.include_picture is True |
| 18 | + assert params.include_chart is True |
| 19 | + assert params.include_formula is True |
| 20 | + |
| 21 | + |
| 22 | +def test_create_task_filtered_params_specific_tasks(): |
| 23 | + """Test with specific task list.""" |
| 24 | + params = create_task_filtered_params(["ocr", "layout"]) |
| 25 | + assert params.include_ocr is True |
| 26 | + assert params.include_layout is True |
| 27 | + assert params.include_otsl is False |
| 28 | + assert params.include_code is False |
| 29 | + assert params.include_picture is False |
| 30 | + assert params.include_chart is False |
| 31 | + assert params.include_formula is False |
| 32 | + |
| 33 | + |
| 34 | +def test_create_task_filtered_params_with_layout(): |
| 35 | + """Test with layout in task list.""" |
| 36 | + params = create_task_filtered_params(["layout"]) |
| 37 | + assert params.include_layout is True |
| 38 | + assert params.add_location is True |
| 39 | + |
| 40 | + |
| 41 | +def test_create_task_filtered_params_without_layout(): |
| 42 | + """Test without layout in task list.""" |
| 43 | + params = create_task_filtered_params(["ocr"]) |
| 44 | + assert params.include_layout is False |
| 45 | + assert params.add_location is False |
| 46 | + |
| 47 | + |
| 48 | +def test_create_task_filtered_params_with_kwargs(): |
| 49 | + """Test with additional kwargs.""" |
| 50 | + params = create_task_filtered_params(["ocr"], xsize=1000, ysize=1000) |
| 51 | + assert params.xsize == 1000 |
| 52 | + assert params.ysize == 1000 |
| 53 | + assert params.include_ocr is True |
| 54 | + |
| 55 | + |
| 56 | +def test_doctags_exclude_ocr(sample_doc): |
| 57 | + """Test excluding OCR.""" |
| 58 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 59 | + serializer.params = serializer.params.merge_with_patch( |
| 60 | + DocTagsParams(include_ocr=False).model_dump() |
| 61 | + ) |
| 62 | + result = serializer.serialize() |
| 63 | + assert result.text is not None |
| 64 | + |
| 65 | + |
| 66 | +def test_doctags_exclude_otsl(sample_doc): |
| 67 | + """Test excluding OTSL.""" |
| 68 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 69 | + serializer.params = serializer.params.merge_with_patch( |
| 70 | + DocTagsParams(include_otsl=False, include_layout=True).model_dump() |
| 71 | + ) |
| 72 | + result = serializer.serialize() |
| 73 | + assert result.text is not None |
| 74 | + |
| 75 | + |
| 76 | +def test_doctags_exclude_picture(sample_doc): |
| 77 | + """Test excluding pictures.""" |
| 78 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 79 | + serializer.params = serializer.params.merge_with_patch( |
| 80 | + DocTagsParams(include_picture=False).model_dump() |
| 81 | + ) |
| 82 | + result = serializer.serialize() |
| 83 | + assert result.text is not None |
| 84 | + |
| 85 | + |
| 86 | +def test_doctags_exclude_chart(sample_doc): |
| 87 | + """Test excluding charts.""" |
| 88 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 89 | + serializer.params = serializer.params.merge_with_patch( |
| 90 | + DocTagsParams(include_chart=False).model_dump() |
| 91 | + ) |
| 92 | + result = serializer.serialize() |
| 93 | + assert result.text is not None |
| 94 | + |
| 95 | + |
| 96 | +def test_doctags_exclude_code(sample_doc): |
| 97 | + """Test excluding code.""" |
| 98 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 99 | + serializer.params = serializer.params.merge_with_patch( |
| 100 | + DocTagsParams(include_code=False).model_dump() |
| 101 | + ) |
| 102 | + result = serializer.serialize() |
| 103 | + assert result.text is not None |
| 104 | + |
| 105 | + |
| 106 | +def test_doctags_exclude_formula(sample_doc): |
| 107 | + """Test excluding formulas.""" |
| 108 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 109 | + serializer.params = serializer.params.merge_with_patch( |
| 110 | + DocTagsParams(include_formula=False).model_dump() |
| 111 | + ) |
| 112 | + result = serializer.serialize() |
| 113 | + assert result.text is not None |
| 114 | + |
| 115 | + |
| 116 | +def test_doctags_no_layout_no_locations(sample_doc): |
| 117 | + """Test no locations when layout is disabled.""" |
| 118 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 119 | + serializer.params = serializer.params.merge_with_patch( |
| 120 | + DocTagsParams(include_layout=False, add_location=True).model_dump() |
| 121 | + ) |
| 122 | + result = serializer.serialize() |
| 123 | + assert "<loc" not in result.text or result.text == "" |
| 124 | + |
| 125 | + |
| 126 | +def test_doctags_layout_with_locations(sample_doc): |
| 127 | + """Test locations when layout is enabled.""" |
| 128 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 129 | + serializer.params = serializer.params.merge_with_patch( |
| 130 | + DocTagsParams(include_layout=True, add_location=True).model_dump() |
| 131 | + ) |
| 132 | + result = serializer.serialize() |
| 133 | + assert result.text is not None |
| 134 | + |
| 135 | + |
| 136 | +def test_doctags_table_location_without_otsl(sample_doc): |
| 137 | + """Test table locations without OTSL.""" |
| 138 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 139 | + serializer.params = serializer.params.merge_with_patch( |
| 140 | + DocTagsParams( |
| 141 | + include_otsl=False, include_layout=True, add_location=True |
| 142 | + ).model_dump() |
| 143 | + ) |
| 144 | + result = serializer.serialize() |
| 145 | + assert result.text is not None |
| 146 | + |
| 147 | + |
| 148 | +def test_doctags_table_caption_without_otsl(sample_doc): |
| 149 | + """Test table captions without OTSL.""" |
| 150 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 151 | + serializer.params = serializer.params.merge_with_patch( |
| 152 | + DocTagsParams( |
| 153 | + include_otsl=False, include_layout=True, add_caption=True |
| 154 | + ).model_dump() |
| 155 | + ) |
| 156 | + result = serializer.serialize() |
| 157 | + assert result.text is not None |
| 158 | + |
| 159 | + |
| 160 | +def test_doctags_multiple_filters(sample_doc): |
| 161 | + """Test multiple filters.""" |
| 162 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 163 | + serializer.params = serializer.params.merge_with_patch( |
| 164 | + DocTagsParams( |
| 165 | + include_ocr=True, |
| 166 | + include_otsl=False, |
| 167 | + include_picture=False, |
| 168 | + include_chart=False, |
| 169 | + include_code=False, |
| 170 | + include_formula=False, |
| 171 | + ).model_dump() |
| 172 | + ) |
| 173 | + result = serializer.serialize() |
| 174 | + assert result.text is not None |
| 175 | + |
| 176 | + |
| 177 | +def test_doctags_layout_mode_only(sample_doc): |
| 178 | + """Test layout mode only.""" |
| 179 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 180 | + serializer.params = serializer.params.merge_with_patch( |
| 181 | + DocTagsParams( |
| 182 | + include_layout=True, |
| 183 | + layout_mode_only=True, |
| 184 | + add_location=True, |
| 185 | + add_content=False, |
| 186 | + ).model_dump() |
| 187 | + ) |
| 188 | + result = serializer.serialize() |
| 189 | + assert result.text is not None |
| 190 | + |
| 191 | + |
| 192 | +def test_doctags_params_mode_minified(sample_doc): |
| 193 | + """Test minified mode.""" |
| 194 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 195 | + serializer.params = serializer.params.merge_with_patch( |
| 196 | + DocTagsParams(mode=DocTagsParams.Mode.MINIFIED).model_dump() |
| 197 | + ) |
| 198 | + result = serializer.serialize() |
| 199 | + assert result.text is not None |
| 200 | + assert serializer.params.mode == DocTagsParams.Mode.MINIFIED |
| 201 | + |
| 202 | + |
| 203 | +def test_doctags_params_mode_human_friendly(sample_doc): |
| 204 | + """Test human-friendly mode.""" |
| 205 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 206 | + serializer.params = serializer.params.merge_with_patch( |
| 207 | + DocTagsParams(mode=DocTagsParams.Mode.HUMAN_FRIENDLY).model_dump() |
| 208 | + ) |
| 209 | + result = serializer.serialize() |
| 210 | + assert result.text is not None |
| 211 | + |
| 212 | + |
| 213 | +def test_doctags_picture_serializer_helper_methods(sample_doc): |
| 214 | + """Test picture serializer helper methods.""" |
| 215 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 216 | + picture_serializer = serializer.picture_serializer |
| 217 | + |
| 218 | + assert hasattr(picture_serializer, "_get_predicted_class") |
| 219 | + assert hasattr(picture_serializer, "_is_chart_type") |
| 220 | + assert hasattr(picture_serializer, "_get_molecule_smi") |
| 221 | + assert hasattr(picture_serializer, "_get_tabular_chart_data") |
| 222 | + assert hasattr(picture_serializer, "_build_body_content") |
| 223 | + |
| 224 | + |
| 225 | +def test_doctags_should_process_item_logic(sample_doc): |
| 226 | + """Test item processing logic.""" |
| 227 | + serializer = DocTagsDocSerializer(doc=sample_doc) |
| 228 | + |
| 229 | + serializer.params = serializer.params.merge_with_patch( |
| 230 | + DocTagsParams(include_layout=True).model_dump() |
| 231 | + ) |
| 232 | + result = serializer.serialize() |
| 233 | + assert result.text is not None |
| 234 | + |
| 235 | + serializer.params = serializer.params.merge_with_patch( |
| 236 | + DocTagsParams(include_layout=True, include_otsl=False).model_dump() |
| 237 | + ) |
| 238 | + result = serializer.serialize() |
| 239 | + assert result.text is not None |
0 commit comments