Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

feat: (experimental) introduce new document format #21

Merged
merged 43 commits into from
Oct 1, 2024
Merged

feat: (experimental) introduce new document format #21

merged 43 commits into from
Oct 1, 2024

Conversation

cau-git
Copy link
Contributor

@cau-git cau-git commented Sep 17, 2024

No description provided.

@cau-git cau-git force-pushed the cau/new-format-dev branch 2 times, most recently from 334e1b7 to 5df614b Compare September 17, 2024 13:23
@cau-git cau-git changed the title Draft new docling document format, pydantic model and tests feat: Draft new docling document format, pydantic model and tests Sep 17, 2024
@cau-git
Fix tests to have unique document_hashes per test
43c23b9 
Signed-off-by: Christoph Auer <[email protected]>
@cau-git
Manual update from main
384d12a 
Signed-off-by: Christoph Auer <[email protected]>
@cau-git
Merge from main
ed08764 
Signed-off-by: Christoph Auer <[email protected]>
@cau-git
Merge from main
1c753e9 
Signed-off-by: Christoph Auer <[email protected]>
@cau-git
Move new-format to experimental path
3d3c2f3 
Signed-off-by: Christoph Auer <[email protected]>
@cau-git
Move new-format to experimental path
bdbd93e 
Signed-off-by: Christoph Auer <[email protected]>
@cau-git
Turn captions into list field
f791f74 
Signed-off-by: Christoph Auer <[email protected]>
cau-git and others added 4 commits September 27, 2024 15:17
@cau-git
Remove hash, renamings
23e1a52 
Signed-off-by: Christoph Auer <[email protected]>
@PeterStaar-IBM
updating the tests
fe72b3b 
Signed-off-by: Peter Staar <[email protected]>
@PeterStaar-IBM
fixed merge conflicts
b6a7049 
Signed-off-by: Peter Staar <[email protected]>
@PeterStaar-IBM
added some unit tests for DocItem
baced33 
Signed-off-by: Peter Staar <[email protected]>
PeterStaar-IBM
PeterStaar-IBM reviewed Sep 27, 2024
View reviewed changes
test/test_docling_doc.py
)
from docling_core.types.experimental.labels import DocItemLabel, GroupLabel

def test_docitems():
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@cau-git added a unit-test, so we can detect if new classes appear or we have added the serialisation of a DocItem subclass

maxmnemonic
maxmnemonic reviewed Sep 27, 2024
View reviewed changes
docling_core/types/experimental/document.py

class GroupItem(NodeItem): # Container type, can't be a leaf node
"""GroupItem."""

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Add provenance field to GroupItem or NodeItem,
to support provenance of entire lists, or slides (that are collections of elements)
In case of lists - it might be that individual bounding boxes are not possible or hard to obtain, but bounding box of an entire list is readily available. And in any way we would like to keep bounding boxes of lists, for this Group should support provenance.

ceberam and others added 7 commits September 27, 2024 19:17
PeterStaar-IBM
PeterStaar-IBM previously approved these changes Sep 30, 2024
View reviewed changes
Copy link
Contributor

@PeterStaar-IBM PeterStaar-IBM left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let's merge this!

@cau-git cau-git dismissed PeterStaar-IBM’s stale review September 30, 2024 12:55

Still missing work.

vagenas
vagenas reviewed Sep 30, 2024
View reviewed changes
test/data/experimental/dummy_doc.yaml Outdated
Comment on lines 21 to 42
name: "_root_"
self_ref: "#/body"
parent: null # Only root elements have no parent.
children: # only the first-level children appear here, as references (RefItem)
- $ref: "/texts/1"
- $ref: "/pictures/0"
- $ref: "/texts/3"
- $ref: "/tables/0"

# All groups of items nested deeper in body or furniture roots, type List[GroupItem]
groups: [] # The parent + children relations capture nesting and reading-order.

# All elements that have a text-string representation, type TextItem.
# This is a flat list of all elements without implied order.
texts:
- orig: "arXiv:2206.01062v1 [cs.CV] 2 Jun 2022"
text: "arXiv:2206.01062v1 [cs.CV] 2 Jun 2022"
self_ref: "#/texts/0"
label: "page_header"
parent:
$ref: "#/furniture"
children: []
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@cau-git I see something that may be an inconsistency:

  • the parent has the # prefix (fragment separator for URI fragment): $ref: "#/furniture", while
  • the children do not have this prefix, e.g. - $ref: "/texts/1"

Can you clarify?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Well spotted. The children must have it too.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@vagenas Can you do the update?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The fix needed is only to correct manually this file, since it was hand-crafted.
We should also consider validating the ref property to ensure it always follows the #{}/{} pattern.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@PeterStaar-IBM @cau-git

  1. validation added (which individually as expected causes the tests to fail due to the invalid data) and
  2. test data fixed

vagenas
vagenas reviewed Sep 30, 2024
View reviewed changes
test/test_docling_doc.py
fig_caption = doc.add_text(
label=DocItemLabel.CAPTION, text="This is the caption of figure 1."
)
fig_item = doc.add_picture(data=BasePictureData(), caption=fig_caption)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@cau-git fig_item here is not used.

@PeterStaar-IBM PeterStaar-IBM changed the title feat: Draft new docling document format, pydantic model and tests feat: (experimental) introduce new document format Oct 1, 2024
@cau-git cau-git merged commit 688789e into main Oct 1, 2024
5 checks passed
@cau-git cau-git deleted the cau/new-format-dev branch October 1, 2024 07:41
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@ceberam ceberam ceberam left review comments

@maxmnemonic maxmnemonic maxmnemonic approved these changes

@nikos-livathinos nikos-livathinos nikos-livathinos approved these changes

@dolfim-ibm dolfim-ibm dolfim-ibm approved these changes

@PeterStaar-IBM PeterStaar-IBM PeterStaar-IBM approved these changes

@vagenas vagenas vagenas approved these changes

@nassarofficial nassarofficial Awaiting requested review from nassarofficial

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
7 participants
@cau-git @vagenas @ceberam @PeterStaar-IBM @dolfim-ibm @nikos-livathinos @maxmnemonic