def test_should_find_lines_of_page_with_blocks(self):
lines = [E.TEXT(), E.TEXT()]
page = E.PAGE(E.BLOCK(*lines))
doc = LxmlStructuredDocument(
E.DOCUMENT(
page,
# add another page just for effect
E.PAGE(E.BLOCK(E.TEXT()))))
assert list(doc.get_lines_of_page(page)) == lines
def test_should_not_fail_setting_empty_tag_to_none(self):
token = E.TEXT()
doc = LxmlStructuredDocument(E.DOCUMENT(E.PAGE(E.BLOCK(token))))
doc.set_tag(token, None)
doc.set_tag(token, None, scope=SCOPE_1)
assert doc.get_tag(token) is None
assert doc.get_tag(token, scope=SCOPE_1) is None
def test_should_return_all_tag_by_scope(self):
token = E.TEXT()
doc = LxmlStructuredDocument(E.DOCUMENT(E.PAGE(E.BLOCK(token))))
doc.set_tag(token, TAG_1)
doc.set_tag(token, TAG_2, scope=SCOPE_1)
assert doc.get_tag(token) == TAG_1
assert doc.get_tag(token, scope=SCOPE_1) == TAG_2
assert doc.get_tag_by_scope(token) == {None: TAG_1, SCOPE_1: TAG_2}
def test_should_keep_text_block_structure_with_block(self):
lxml_root = E.DOCUMENT(
E.PAGE(
E.BLOCK(
E.TEXT(
E.TOKEN(
SOME_TEXT,
dict_extend(COMMON_LXML_TOKEN_ATTRIBS,
{LXML.BASE: SOME_BASE}))))))
svg_pages = list(iter_svg_pages_for_lxml(lxml_root))
assert len(svg_pages) == 1
first_page = svg_pages[0]
svg_text = first_page.find('.//' + SVG_TEXT)
assert svg_text is not None
assert svg_text.getparent().tag == SVG_G
assert svg_text.getparent().getparent().tag == SVG_G
assert svg_text.getparent().getparent().getparent().tag == SVG_DOC
def test_should_set_tag_with_level(self):
token = E.TEXT()
doc = LxmlStructuredDocument(E.DOCUMENT(E.PAGE(E.BLOCK(token))))
doc.set_tag(token, TAG_1, level=2)
assert doc.get_tag(token, level=2) == TAG_1
assert doc.get_tag(token) is None
def test_should_set_tag_with_scope(self):
token = E.TEXT()
doc = LxmlStructuredDocument(E.DOCUMENT(E.PAGE(E.BLOCK(token))))
doc.set_tag(token, TAG_1, scope=SCOPE_1)
assert doc.get_tag(token, scope=SCOPE_1) == TAG_1
assert doc.get_tag(token) is None