yujunjun/unstructured
1
0
Fork 0
mirror of https://github.com/Unstructured-IO/unstructured.git synced 2025-12-27 07:03:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix: remove unused ElementMetadata.section (#2921)

Browse Source 
**Summary**
The `.section` field in `ElementMetadata` is dead code, possibly a
remainder from a prior iteration of `partition_epub()`. In any case, it
is not populated by any partitioner. Remove it and any code that uses
it.
This commit is contained in:
Steve Canny 2024-04-22 16:58:17 -07:00 committed by GitHub
parent 305247b4e1
commit 05ff975081
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
11 changed files with 21 additions and 170 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
CHANGELOG.md
View File

@ -1,3 +1,13 @@
## 0.13.4-dev0
### Enhancements
### Features
### Fixes
* **Remove ElementMetadata.section field.**. This field was unused, not populated by any partitioners.
## 0.13.3
### Enhancements

7
docs/source/core/chunking.rst
View File

@ -152,13 +152,6 @@ following behaviors:
``Title`` element would fit in the prior chunk. This implements the first aspect of the "preserve
section boundaries" contract.
- **Detect metadata.section change.** An element with a new value in ``element.metadata.section`` is
considered to start a new section. When a change in this value is encountered a new chunk is
started. This implements the second aspect of preserving section boundaries. This metadata is not
present in all document formats so is not used alone. An element having ``None`` for this metadata
field is considered to be part of the prior section; a section break is only detected on an
explicit change in value.
- **Respect page boundaries.** Page boundaries can optionally also be respected using the
``multipage_sections`` argument. This defaults to ``True`` meaning that a page break does *not*
start a new chunk. Setting this to ``False`` will separate elements that occur on different pages

63
test_unstructured/chunking/test_base.py
View File

@ -17,7 +17,6 @@ from unstructured.chunking.base import (
TextPreChunk,
TextPreChunkAccumulator,
_TextSplitter,
is_in_next_section,
is_on_next_page,
is_title,
)
@ -1514,68 +1513,6 @@ class DescribeTextPreChunkAccumulator:
# ================================================================================================
class Describe_is_in_next_section:
"""Unit-test suite for `unstructured.chunking.base.is_in_next_section()` function.
`is_in_next_section()` is not itself a predicate, rather it returns a predicate on Element
(`Callable[[Element], bool]`) that can be called repeatedly to detect section changes in an
element stream.
"""
def it_is_false_for_the_first_element_when_it_has_a_non_None_section(self):
"""This is an explicit first-section; first-section does not represent a section break."""
pred = is_in_next_section()
assert not pred(Text("abcd", metadata=ElementMetadata(section="Introduction")))
def and_it_is_false_for_the_first_element_when_it_has_a_None_section(self):
"""This is an anonymous first-section; still doesn't represent a section break."""
pred = is_in_next_section()
assert not pred(Text("abcd"))
def it_is_false_for_None_section_elements_that_follow_an_explicit_first_section(self):
"""A `None` section element is considered to continue the prior section."""
pred = is_in_next_section()
assert not pred(Text("abcd", metadata=ElementMetadata(section="Introduction")))
assert not pred(Text("efgh"))
assert not pred(Text("ijkl"))
def and_it_is_false_for_None_section_elements_that_follow_an_anonymous_first_section(self):
"""A `None` section element is considered to continue the prior section."""
pred = is_in_next_section()
assert not pred(Text("abcd"))
assert not pred(Text("efgh"))
assert not pred(Text("ijkl"))
def it_is_false_for_matching_section_elements_that_follow_an_explicit_first_section(self):
pred = is_in_next_section()
assert not pred(Text("abcd", metadata=ElementMetadata(section="Introduction")))
assert not pred(Text("efgh", metadata=ElementMetadata(section="Introduction")))
assert not pred(Text("ijkl", metadata=ElementMetadata(section="Introduction")))
def it_is_true_for_an_explicit_section_element_that_follows_an_anonymous_first_section(self):
pred = is_in_next_section()
assert not pred(Text("abcd"))
assert not pred(Text("efgh"))
assert pred(Text("ijkl", metadata=ElementMetadata(section="Introduction")))
def and_it_is_true_for_a_different_explicit_section_that_follows_an_explicit_section(self):
pred = is_in_next_section()
assert not pred(Text("abcd", metadata=ElementMetadata(section="Introduction")))
assert pred(Text("efgh", metadata=ElementMetadata(section="Summary")))
def it_is_true_whenever_the_section_explicitly_changes_except_at_the_start(self):
pred = is_in_next_section()
assert not pred(Text("abcd"))
assert pred(Text("efgh", metadata=ElementMetadata(section="Introduction")))
assert not pred(Text("ijkl"))
assert not pred(Text("mnop", metadata=ElementMetadata(section="Introduction")))
assert not pred(Text("qrst"))
assert pred(Text("uvwx", metadata=ElementMetadata(section="Summary")))
assert not pred(Text("yzab", metadata=ElementMetadata(section="Summary")))
assert not pred(Text("cdef"))
assert pred(Text("ghij", metadata=ElementMetadata(section="Appendix")))
class Describe_is_on_next_page:
"""Unit-test suite for `unstructured.chunking.base.is_on_next_page()` function.

37
test_unstructured/chunking/test_title.py
View File

@ -139,43 +139,6 @@ def test_chunk_by_title():
)
def test_chunk_by_title_respects_section_change():
elements: list[Element] = [
Title("A Great Day", metadata=ElementMetadata(section="first")),
Text("Today is a great day.", metadata=ElementMetadata(section="second")),
Text("It is sunny outside.", metadata=ElementMetadata(section="second")),
Table("Heading\nCell text"),
Title("An Okay Day"),
Text("Today is an okay day."),
Text("It is rainy outside."),
Title("A Bad Day"),
Text(
"Today is a bad day.",
metadata=ElementMetadata(
regex_metadata={"a": [RegexMetadata(text="A", start=0, end=1)]},
),
),
Text("It is storming outside."),
CheckBox(),
]
chunks = chunk_by_title(elements, combine_text_under_n_chars=0)
assert chunks == [
CompositeElement(
"A Great Day",
),
CompositeElement(
"Today is a great day.\n\nIt is sunny outside.",
),
Table("Heading\nCell text"),
CompositeElement("An Okay Day\n\nToday is an okay day.\n\nIt is rainy outside."),
CompositeElement(
"A Bad Day\n\nToday is a bad day.\n\nIt is storming outside.",
),
]
def test_chunk_by_title_separates_by_page_number():
elements: list[Element] = [
Title("A Great Day", metadata=ElementMetadata(page_number=1)),

2
test_unstructured/partition/epub/test_epub.py
View File

@ -77,7 +77,6 @@ def test_partition_epub_from_filename_exclude_metadata():
assert elements[0].metadata.filetype is None
assert elements[0].metadata.page_name is None
assert elements[0].metadata.filename is None
assert elements[0].metadata.section is None
def test_partition_epub_from_file_exlcude_metadata():
@ -87,7 +86,6 @@ def test_partition_epub_from_file_exlcude_metadata():
assert elements[0].metadata.filetype is None
assert elements[0].metadata.page_name is None
assert elements[0].metadata.filename is None
assert elements[0].metadata.section is None
def test_partition_epub_metadata_date(

3
test_unstructured/staging/test_base.py
View File

@ -166,7 +166,6 @@ def test_default_pandas_dtypes():
sent_from=["sent", "from"],
sent_to=["sent", "to"],
subject="subject",
section="section",
header_footer_type="header_footer_type",
emphasized_text_contents=["emphasized", "text", "contents"],
emphasized_text_tags=["emphasized", "text", "tags"],
@ -321,7 +320,6 @@ def test_convert_to_coco():
sent_from=["sent", "from"],
sent_to=["sent", "to"],
subject="subject",
section="section",
header_footer_type="header_footer_type",
emphasized_text_contents=["emphasized", "text", "contents"],
emphasized_text_tags=["emphasized", "text", "tags"],
@ -366,7 +364,6 @@ def test_convert_to_coco():
sent_from=["sent", "from"],
sent_to=["sent", "to"],
subject="subject",
section="section",
header_footer_type="header_footer_type",
emphasized_text_contents=["emphasized", "text", "contents"],
emphasized_text_tags=["emphasized", "text", "tags"],

2
unstructured/__version__.py
View File

@ -1 +1 @@
__version__ = "0.13.3" # pragma: no cover
__version__ = "0.13.4-dev0" # pragma: no cover

45
unstructured/chunking/base.py
View File

@ -1022,51 +1022,6 @@ class TextPreChunkAccumulator:
# ================================================================================================
def is_in_next_section() -> BoundaryPredicate:
"""Not a predicate itself, calling this returns a predicate that triggers on each new section.
The lifetime of the returned callable cannot extend beyond a single element-stream because it
stores current state (current section) that is particular to that element stream.
A "section" of this type is particular to the EPUB format (so far) and not to be confused with
a "section" composed of a section-heading (`Title` element) followed by content elements.
The returned predicate tracks the current section, starting at `None`. Calling with an element
with a different value for `metadata.section` returns True, indicating the element starts a new
section boundary, and updates the enclosed section name ready for the next transition.
"""
current_section: Optional[str] = None
is_first: bool = True
def section_changed(element: Element) -> bool:
nonlocal current_section, is_first
section = element.metadata.section
# -- The first element never reports a section break, it starts the first section of the
# -- document. That section could be named (section is non-None) or anonymous (section is
# -- None). We don't really have to care.
if is_first:
current_section = section
is_first = False
return False
# -- An element with a `None` section is assumed to continue the current section. It never
# -- updates the current-section because once set, the current-section is "sticky" until
# -- replaced by another explicit section.
if section is None:
return False
# -- another element with the same section continues that section --
if section == current_section:
return False
current_section = section
return True
return section_changed
def is_on_next_page() -> BoundaryPredicate:
"""Not a predicate itself, calling this returns a predicate that triggers on each new page.

2
unstructured/chunking/title.py
View File

@ -13,7 +13,6 @@ from unstructured.chunking.base import (
ChunkingOptions,
PreChunkCombiner,
PreChunker,
is_in_next_section,
is_on_next_page,
is_title,
)
@ -121,7 +120,6 @@ class _ByTitleChunkingOptions(ChunkingOptions):
def iter_boundary_predicates() -> Iterator[BoundaryPredicate]:
yield is_title
yield is_in_next_section()
if not self.multipage_sections:
yield is_on_next_page()

18
unstructured/documents/elements.py
View File

@ -191,8 +191,6 @@ class ElementMetadata:
parent_id: Optional[str]
# -- "fields" e.g. status, dept.no, etc. extracted from text via regex --
regex_metadata: Optional[dict[str, list[RegexMetadata]]]
# -- EPUB document section --
section: Optional[str]
# -- e-mail specific metadata fields --
sent_from: Optional[list[str]]
@ -235,7 +233,6 @@ class ElementMetadata:
page_number: Optional[int] = None,
parent_id: Optional[str] = None,
regex_metadata: Optional[dict[str, list[RegexMetadata]]] = None,
section: Optional[str] = None,
sent_from: Optional[list[str]] = None,
sent_to: Optional[list[str]] = None,
signature: Optional[str] = None,
@ -275,7 +272,6 @@ class ElementMetadata:
self.page_number = page_number
self.parent_id = parent_id
self.regex_metadata = regex_metadata
self.section = section
self.sent_from = sent_from
self.sent_to = sent_to
self.signature = signature
@ -488,7 +484,6 @@ class ConsolidationStrategy(enum.Enum):
"page_number": cls.FIRST,
"parent_id": cls.DROP,
"regex_metadata": cls.REGEX,
"section": cls.FIRST,
"sent_from": cls.FIRST,
"sent_to": cls.FIRST,
"signature": cls.FIRST,
@ -671,7 +666,7 @@ class ElementType:
class Element(abc.ABC):
"""An element is a section of a page in the document.
"""An element is a semantically-coherent component of a document, often a paragraph.
There are a few design principles that are followed when creating an element:
1. It will always have an ID, which by default is a random UUID.
@ -694,7 +689,9 @@ class Element(abc.ABC):
metadata: Optional[ElementMetadata] = None,
detection_origin: Optional[str] = None,
):
if element_id is not None and not isinstance(element_id, str):
if element_id is not None and not isinstance(
element_id, str
): # pyright: ignore[reportUnnecessaryIsInstance]
raise ValueError("element_id must be of type str or None.")
self._element_id = element_id
@ -885,7 +882,12 @@ class Formula(Text):
class CompositeElement(Text):
"""A section of text consisting of a combination of elements."""
"""A chunk formed from text (non-Table) elements.
Only produced by chunking. An instance may be formed by combining one or more sequential
elements produced by partitioning. It it also used when text-splitting an "oversized" element,
a single element that by itself is larger than the requested chunk size.
"""
category = "CompositeElement"

2
unstructured/partition/common.py
View File

@ -272,7 +272,6 @@ def add_element_metadata(
text_as_html: Optional[str] = None,
coordinates: Optional[tuple[tuple[float, float], ...]] = None,
coordinate_system: Optional[CoordinateSystem] = None,
section: Optional[str] = None,
image_path: Optional[str] = None,
detection_origin: Optional[str] = None,
languages: Optional[List[str]] = None,
@ -324,7 +323,6 @@ def add_element_metadata(
link_start_indexes=link_start_indexes,
emphasized_text_contents=emphasized_text_contents,
emphasized_text_tags=emphasized_text_tags,
section=section,
category_depth=depth,
image_path=image_path,
languages=languages,