In DOCX, like HTML, a table cell can itself contain a table. This is not
uncommon and is typically used for formatting purposes.
When a DOCX table is nested, create nested HTML tables to reflect that
structure and create a plain-text table with captures all the text in
nested tables, formatting it as a reasonable facsimile of a table.
This implements the solution described and spiked in PR #1952.
---------
Co-authored-by: Bruno Bornsztein <bruno.bornsztein@gmail.com>
Intermittently the various destination test will fail with:
```
{noformat}--- Cleanup done ---
gs://utic-test-ingest-fixtures-output/1699377964/example-docs/
deleting gs://utic-test-ingest-fixtures-output/1699377964
Removing objects:
ERROR: (gcloud.storage.rm) The following URLs matched no objects or files:
-gs://utic-test-ingest-fixtures-output/1699377964
Last ran script: gcs.sh
Error: Process completed with exit code 1.{noformat}
```
Reference trace
[here](https://github.com/Unstructured-IO/unstructured/actions/runs/6787927424/job/18452240764?pr=2020)
After some investigation it looks like this error is due to collisions
that occur because we’re assuming 1s date accuracy is sufficient when
generating (and deleting) "unique" test destination location names. The
likelihood is actually pretty high given that we run these tests against
a test matrix.
Instead we should just use a uuid for these unique destinations.
## Changes
- Use uuidgen instead of `date +%s` for unique destinations
### Summary
Click decorated functions cannot (properly) be called outside of the
click interface. This makes it difficult to reuse the setup
functionality in measure_text_edit_distance or
measure_element_type_accuracy. This PR removes the click decoration and
separates it into a wrapper function purely to execute the command.
### Technical Details
- Changed as suggested in [this StackOverflow
post](https://stackoverflow.com/questions/40091347/call-another-click-command-from-a-click-command)
response
- The locations of these now distinct functions are separate: the
`_command` click-decorated functions stay in ingest/evaluate.py, and the
core functions measure_text_edit_distance and
measure_element_type_accuracy are moved into the unstructured/metrics/
folder (which is a more logical location for them).
- Initial test added for measure_text_edit_distance
### Test
`sh ./test_unstructured_ingest/evaluation-metrics.sh text-extraction`
functionality is unchanged.
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: shreyanid <shreyanid@users.noreply.github.com>
Co-authored-by: Trevor Bossert <37596773+tabossert@users.noreply.github.com>
Summary:
Close: https://github.com/Unstructured-IO/unstructured/issues/1920
* stop passing in empty string from `languages` to tesseract, which will
result in passing empty string to language config `-l` for the tesseract
CLI
* also stop passing in duplicate language code from `languages` to
tesseract OCR
* if we failed to convert any iso languages from the `languages`
parameter, proceed OCR with `eng` as default
### Test
* First confirm the tesseract error `Estimating resolution as X` before
this:
* on the `unstructured-api` repo with main branch, run `make
run-web-app`
* curl to test error from empty string, or just any wrong input like `-F
'languages="eng,de"'`:
```
curl -X 'POST' 'http://0.0.0.0:8000/general/v0/general' \
-H 'accept: application/json' \
-H 'Content-Type: multipart/form-data' \
-F 'files=@sample-docs/layout-parser-paper-with-table.jpg' \
-F 'languages=""' \
-F 'strategy=hi_res' \
-F 'pdf_infer_table_structure=True' \
| jq -C . | less -R
```
* after this change:
* in your unstructured API env, cd to unstructured repo and install it locally with `pip install -e .`
* check out to this branch
* run `make run-web-app` again in api repo
* the curl command return output and see warning in log
---------
Co-authored-by: qued <64741807+qued@users.noreply.github.com>
Per @tabossert we're now using a link shortener behind which we can
rotate the link to keep it current. That way we (🤞 ) never have to
update this here again.
#### Testing:
Links should work. No more links should exist in the documentation
except this one.
Closes#1782
This PR:
- Extends ingest pipeline so that it is possible to select an embedding
provider from a range of providers
- Modifies the ingest embedding test to be a diff test, since the
embedding vectors are reproducible after supporting multiple providers
Additional info on the chosen provider for the test:
- Found `langchain.embeddings.HuggingFaceEmbeddings` to be deterministic
even when there's no seed set
- Took 6.84s to pass a unit test with the provider (without cache,
including model download)
- `langchain.embeddings.HuggingFaceEmbeddings` runs in local, making it
zero cost
For all these reasons, testing embedding modules with the Huggingface
model seems to be making sense
---------
Co-authored-by: cragwolfe <crag@unstructured.io>
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: ahmetmeleq <ahmetmeleq@users.noreply.github.com>
We currently have a method to trigger the ingest fixture workflow by
commit message in addition to workflow dispatch (the trigger in gha
gui). The former requires that the workflow run on every push. Because
nobody uses the former, let's scrap it and save the time in CI.
### Description
* A full schema was introduced to map the type of all output content
from the json partition output and mapped to a flattened table structure
to leverage table-based destination connectors. The delta table
destination connector was updated at the moment to take advantage of
this.
* Existing method to convert to a dataframe was updated because it had a
bug in it. Object content in the metadata would have the key name
changed when flattened but then this would be omitted since it didn't
exist in the `_get_metadata_table_fieldnames` response.
* Unit test was added to make sure we handle all values possible in an
Element when converting to a table
* Delta table ingest test was split into a source and destination test
(looking ahead to split these up in CI)
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: rbiseck3 <rbiseck3@users.noreply.github.com>
### Description
Update all destination tests to match pattern:
* Don't omit any metadata to check full schema
* Move azure cognitive dest test from src to dest
* Split delta table test into seperate src and dest tests
* Fix azure cognitive search and add to dest tests being run (wasn't
being run originally)
Remove bullets not related to end-user consumption of the unstructured
library.
Co-authored-by: shreyanid <42684285+shreyanid@users.noreply.github.com>
This PR resolves
[CORE-2453](https://unstructured-ai.atlassian.net/browse/CORE-2453):
- parametrizes the output folder so that ingest output files can be
saved other than the same place where the scripts are; this is set by
env `OUTPUT_ROOT`
- parametrize the python path `PYTHONPATH` to first check existing
definition before default to `.`, the current folder
- parametrize the run script that carries out ingest using `RUN_SCRIPT`,
default is still `./unstructured/ingest/main.py`
These changes allows us to run ingest test with more control. To test:
- run `OUTPUT_ROOT=/tmp
./test_unstructured_ingest/src/local-single-file.sh`: the output now
should be in `/tmp` instead of in the ingest test folder
- run `RUN_SCRIPT=/hope/you/do/not/have/this/folder
./test_unstructured_ingest/src/local-single-file.sh` would raise an
error because system can't find `/hope/you/do/not/have/this/folder`
- run `RUN_SCRIPT=./unstructured/ingest/main.py
./test_unstructured_ingest/src/local-single-file.sh` should run as
normal
- do the following
```bash
cp ./unstructured/ingest/main.py /tmp/main.py
OUTPUT_ROOT=/tmp PYTHONPATH=$(pwd) RUN_SCRIPT=./unstructured/ingest/main.py ./test_unstructured_ingest/src/local-single-file.sh
```
This will run and generate output at `/tmp`
[CORE-2453]:
https://unstructured-ai.atlassian.net/browse/CORE-2453?atlOrigin=eyJpIjoiNWRkNTljNzYxNjVmNDY3MDlhMDU5Y2ZhYzA5YTRkZjUiLCJwIjoiZ2l0aHViLWNvbS1KU1cifQ
### Description
To always support the latest changed to the partition method and the
possible kwargs it supports, the ingest CLI has been refactored to take
in a valid json string to represent those values to allow a user more
flexibility with controlling the partition method.
### Summary
To combine ingest and holistic metrics efforts, add the `doctype` field
to the results from the functions in evaluate.py for use in subsequent
aggregation functions.
### Test
Run `sh ./test_unstructured_ingest/evaluation-metrics.sh
text-extraction` and there will be a new doctype column with the file's
doctype extension.
<img width="508" alt="Screenshot 2023-11-01 at 2 23 11 PM"
src="https://github.com/Unstructured-IO/unstructured/assets/42684285/44583da9-e7ef-4142-be72-c2247b954bcf">
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: shreyanid <shreyanid@users.noreply.github.com>
- This PR adds a function to check if a piece of text only contains a
bullet (no text) to prevent creating an empty element.
- Also fixed a test that had a typo.
This PR resolves#1294 by adding a Makefile to compile requirements.
This makefile respects the dependencies between file and will compile
them in order. E.g., extra-*.txt will be compiled __after__ base.txt is
updated.
Test locally by simply running `make pip-compile` or `cd requirements &&
make clean && make all`
---------
Co-authored-by: qued <64741807+qued@users.noreply.github.com>
### Summary
We no longer use the "bricks" terminology for partioning functions, etc
in the library. This PR updates various references to bricks within the
repo and the docs. This is just an initial pass to swap the terminology
out, it'll likely be helpful to reorganize the docs a bit as well.
---------
Co-authored-by: qued <64741807+qued@users.noreply.github.com>
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
*Reviewer:* May be quicker to review commit by commit as they are quite
distinct and well-groomed to each focus on a single clean-up task.
Clean up odds-and-ends in the docx partitioner in preparation for adding
nested-tables support in a closely following PR.
1. Remove obsolete TODOs now in GitHub issues, which is probably where
they belong in future anyway.
2. Remove local DOCX "workaround" code that has been implemented
upstream and is now obsolete.
3. "Clean" the docx tests, introducing strict typing, extracting a
fixture or two, and generally tightening things up.
4. Extract docx-local versions of
`unstructured.partition.common.convert_ms_office_table_to_text()` which
will be the base for adding nested-table support. More information on
why this is required in that commit.
**Executive Summary.** When the elements in a _section_ are combined
into a _chunk_, the metadata in each of the elements is _consolidated_
into a single `ElementMetadata` instance. There are two main problems
with the current implementation:
1. The current algorithm simply uses the metadata of the first element
as the metadata for the chunk. This produces:
- **empty chunk metadata** when the first element has no metadata, such
as a `PageBreak("")`
- **missing chunk metadata** when the first element contains only
partial metadata such as a `Header()` or `Footer()`
- **misleading metadata** when the first element contains values
applicable only to that element, such as `category_depth`, `coordinates`
(bounding-box), `header_footer_type`, or `parent_id`
2. Second, list metadata such as `emphasized_text_content`,
`emphasized_text_tags`, `link_texts` and `link_urls` is only combined
when it is unique within the combined list. These lists are "unzipped"
pairs. For example, the first `link_texts` corresponds to the first
`link_urls` value. When an item is removed from one (because it matches
a prior entry) and not the other (say same text "here" but different
URL) the positional correspondence is broken and downstream processing
will at best be wrong, at worst raise an exception.
### Technical Discussion
Element metadata cannot be determined in the general case simply by
sampling that of the first element. At the same time, a simple union of
all values is also not sufficient. To effectively consolidate the
current variety of metadata fields we need four distinct strategies,
selecting which to apply to each field based on that fields provenance
and other characteristics.
The four strategies are:
- `FIRST` - Select the first non-`None` value across all the elements.
Several fields are determined by the document source (`filename`,
`file_directory`, etc.) and will not change within the output of a
single partitioning run. They might not appear in every element, but
they will be the same whenever they do appear. This strategy takes the
first one that appears, if any, as proxy for the value for the entire
chunk.
- `LIST` - Consolidate the four list fields like
`emphasized_text_content` and `link_urls` by concatenating them in
element order (no set semantics apply). All values from `elements[n]`
appear before those from `elements[n+1]` and existing order is
preserved.
- `LIST_UNIQUE` - Combine only unique elements across the (list) values
of the elements, preserving order in which a unique item first appeared.
- `REGEX` - Regex metadata has its own rules, including adjusting the
`start` and `end` offset of each match based its new position in the
concatenated text.
- `DROP` - Not all metadata can or should appear in a chunk. For
example, a chunk cannot be guaranteed to have a single `category_depth`
or `parent_id`.
Other strategies such as `COORDINATES` could be added to consolidate the
bounding box of the chunk from the coordinates of its elements, roughly
`min(lefts)`, `max(rights)`, etc. Others could be `LAST`, `MAJORITY`, or
`SUM` depending on how metadata evolves.
The proposed strategy assignments are these:
- `attached_to_filename`: FIRST,
- `category_depth`: DROP,
- `coordinates`: DROP,
- `data_source`: FIRST,
- `detection_class_prob`: DROP, # -- ? confirm --
- `detection_origin`: DROP, # -- ? confirm --
- `emphasized_text_contents`: LIST,
- `emphasized_text_tags`: LIST,
- `file_directory`: FIRST,
- `filename`: FIRST,
- `filetype`: FIRST,
- `header_footer_type`: DROP,
- `image_path`: DROP,
- `is_continuation`: DROP, # -- not expected, added by chunking, not
before --
- `languages`: LIST_UNIQUE,
- `last_modified`: FIRST,
- `link_texts`: LIST,
- `link_urls`: LIST,
- `links`: DROP, # -- deprecated field --
- `max_characters`: DROP, # -- unused in code, probably remove from
ElementMetadata --
- `page_name`: FIRST,
- `page_number`: FIRST,
- `parent_id`: DROP,
- `regex_metadata`: REGEX,
- `section`: FIRST, # -- section unconditionally breaks on new section
--
- `sent_from`: FIRST,
- `sent_to`: FIRST,
- `subject`: FIRST,
- `text_as_html`: DROP, # -- not expected, only occurs in TableSection
--
- `url`: FIRST,
**Assumptions:**
- each .eml file is partitioned->chunked separately (not in batches),
therefore
sent-from, sent-to, and subject will not change within a section.
### Implementation
Implementation of this behavior requires two steps:
1. **Collect** all non-`None` values from all elements, each in a
sequence by field-name. Fields not populated in any of the elements do
not appear in the collection.
```python
all_meta = {
"filename": ["memo.docx", "memo.docx"]
"link_texts": [["here", "here"], ["and here"]]
"parent_id": ["f273a7cb", "808b4ced"]
}
```
2. **Apply** the specified strategy to each item in the overall
collection to produce the consolidated chunk meta (see implementation).
### Factoring
For the following reasons, the implementation of metadata consolidation
is extracted from its current location in `chunk_by_title()` to a
handful of collaborating methods in `_TextSection`.
- The current implementation of metadata consolidation "inline" in
`chunk_by_title()` already has too many moving pieces to be understood
without extended study. Adding strategies to that would make it worse.
- `_TextSection` is the only section type where metadata is consolidated
(the other two types always have exactly one element so already exactly
one metadata.)
- `_TextSection` is already the expert on all the information required
to consolidate metadata, in particular the elements that make up the
section and their text.
Some other problems were also fixed in that transition, such as mutation
of elements during the consolidation process.
### Technical Risk: adding new `ElementMetadata` field breaks metadata
If each metadata field requires a strategy assignment to be consolidated
and a developer adds a new `ElementMetadata` field without adding a
corresponding strategy mapping, metadata consolidation could break or
produce incorrect results.
This risk can be mitigated multiple ways:
1. Add a test that verifies a strategy is defined for each
(Recommended).
2. Define a default strategy, either `DROP` or `FIRST` for scalar types,
`LIST` for list types.
3. Raise an exception when an unknown metadata field is encountered.
This PR implements option 1 such that a developer will be notified
before merge if they add a new metadata field but do not define a
strategy for it.
### Other Considerations
- If end-users can in-future add arbitrary metadata fields _before_
chunking, then we'll need to define metadata-consolidation behavior for
such fields. Depending on how we implement user-defined metadata fields
we might:
- Require explicit definition of a new metadata field before use,
perhaps with a method like `ElementMetadata.add_custom_field()` which
requires a consolidation strategy to be defined (and/or has a default
value).
- Have a default strategy, perhaps `DROP` or `FIRST`, or `LIST` if the
field is type `list`.
### Further Context
Metadata is only consolidated for `TextSection` because the other two
section types (`TableSection` and `NonTextSection`) can only contain a
single element.
---
## Further discussion on consolidation strategy by field
### document-static
These fields are very likely to be the same for all elements in a single
document:
- `attached_to_filename`
- `data_source`
- `file_directory`
- `filename`
- `filetype`
- `last_modified`
- `sent_from`
- `sent_to`
- `subject`
- `url`
*Consolidation strategy:* `FIRST` - use first one found, if any.
### section-static
These fields are very likely to be the same for all elements in a single
section, which is the scope we really care about for metadata
consolidation:
- `section` - an EPUB document-section unconditionally starts new
section.
*Consolidation strategy:* `FIRST` - use first one found, if any.
### consolidated list-items
These `List` fields are consolidated by concatenating the lists from
each element that has one:
- `emphasized_text_contents`
- `emphasized_text_tags`
- `link_texts`
- `link_urls`
- `regex_metadata` - special case, this one gets indexes adjusted too.
*Consolidation strategy:* `LIST` - concatenate lists across elements.
### dynamic
These fields are likely to hold unique data for each element:
- `category_depth`
- `coordinates`
- `image_path`
- `parent_id`
*Consolidation strategy:*
- `DROP` as likely misleading.
- `COORDINATES` strategy could be added to compute the bounding box from
all bounding boxes.
- Consider allowing if they are all the same, perhaps an `ALL` strategy.
### slow-changing
These fields are somewhere in-between, likely to be common between
multiple elements but varied within a document:
- `header_footer_type` - *strategy:* drop as not-consolidatable
- `languages` - *strategy:* take first occurence
- `page_name` - *strategy:* take first occurence
- `page_number` - *strategy:* take first occurence, will all be the same
when `multipage_sections` is `False`. Worst-case semantics are "this
chunk began on this page".
### N/A
These field types do not figure in metadata-consolidation:
- `detection_class_prob` - I'm thinking this is for debug and should not
appear in chunks, but need confirmation.
- `detection_origin` - for debug only
- `is_continuation` - is _produced_ by chunking, never by partitioning
(not in our code anyway).
- `links` (deprecated, probably should be dropped)
- `max_characters` - is unused as far as I can tell, is unreferenced in
source code. Should be removed from `ElementMetadata` as far as I can
tell.
- `text_as_html` - only appears in a `Table` element, each of which
appears in its own section so needs no consolidation. Never appears in
`TextSection`.
*Consolidation strategy:* `DROP` any that appear (several never will)
### Summary
Closes#1520
Partial solution to #1521
- Adds an abstraction layer between the user API and the partitioner
implementation
- Adds comments explaining paragraph chunking
- Makes edits to pass strict type-checking for both text.py and
test_text.py
Closes#1870
Defining both `languages` and `ocr_languages` raises a ValueError, but
the api defaults to `ocr_languages` being an empty string, so if users
define `languages` they are automatically hitting the ValueError.
This fix checks if `ocr_languages` is an empty string and converts it to
`None` to avoid this.
### Testing
On the main branch, the following will raise the ValueError, but it will
correctly partition on this branch
```
from unstructured.partition.auto import partition
filename = "example-docs/category-level.docx"
elements = partition(filename,languages=['spa'],ocr_languages="")
elements[0].metadata.languages
```
---------
Co-authored-by: yuming <305248291@qq.com>
Co-authored-by: Yuming Long <63475068+yuming-long@users.noreply.github.com>
Co-authored-by: Austin Walker <awalk89@gmail.com>
### Description
This splits the source ingest tests from the destination ingest tests
since they share a different pattern:
* src tests pull data from a source and compare the partitioned content
to the expected results
* destingation tests leverage the local connector to produce results to
push to a destination and leverages overhead to create temporary
locations at those destinations to write to and delete when done.
Only the src tests create partitioned content that needs to be checked
so the update ingest test CI job only needs to run these.
Refactor the evaluation scripts including
`unstructured/ingest/evaluation.py`
`test_unstructured_ingest/evaluation-metrics.sh` for more structured
code and usage.
- The script is now only use one python script call with param
- Adds function to build string for output_args (`--output_dir
--output_list) and source_args (`--source_dir --source_args`)
- Now accepts evaluation to call as a param, currently only accepts
`text-extraction` and `element-type`
Example to call the function:
```sh evaluation-metrics.sh text-extraction```
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: Klaijan <Klaijan@users.noreply.github.com>
### Description
### Google Drive
The existing service account parameter was expanded to support either a
file path or a json value to generate the credentials when instantiating
the google drive client.
### GCS
Google Cloud Storage already supports the value being passed in, from
their docstring:
> - you may supply a token generated by the
[gcloud](https://cloud.google.com/sdk/docs/)
utility; this is either a python dictionary, the name of a file
containing the JSON returned by logging in with the gcloud CLI tool,
or a Credentials object.
I tested this locally:
```python
from gcsfs import GCSFileSystem
import json
with open("/Users/romanisecke/.ssh/google-cloud-unstructured-ingest-test-d4fc30286d9d.json") as json_file:
json_data = json.load(json_file)
print(json_data)
fs = GCSFileSystem(token=json_data)
print(fs.ls(path="gs://utic-test-ingest-fixtures/"))
```
`['utic-test-ingest-fixtures/ideas-page.html',
'utic-test-ingest-fixtures/nested-1',
'utic-test-ingest-fixtures/nested-2']`
### Description
Add in the fsspec configs needed for the fsspec-based connectors
To match the behavior of the original CLI, the default used by the click
option was mirrored in the base config for the api endpoint.
Closes: #1891 (check the issue for more info)
---------
Co-authored-by: ryannikolaidis <1208590+ryannikolaidis@users.noreply.github.com>
Co-authored-by: ahmetmeleq <ahmetmeleq@users.noreply.github.com>
Co-authored-by: Yao You <yao@unstructured.io>
Using `actions/cache@v3` instead of `actions/cache/restore@v3` for the
cache lookup in `setup_ingest` is causing CI to save the cache twice
when there's a cache miss. This is unnecessary, but I'm also a little
concerned it's causing some sort of race condition since I've seen
instances of CI failing to save due to the cache already existing (which
shouldn't be the case on a cache miss).
This PR switches the lookup to a `restore` action to avoid duplicate
ingest cache saving.
#### Testing:
There should only be one "Post Run actions/cache@v3" step in each
`setup_ingest` job when there's a cache miss.
[Here](https://github.com/Unstructured-IO/unstructured/actions/runs/6707740917/job/18227300507)
is an example of a cache miss running with this PR.
### Description
All http calls being made by the ingest source connectors have been
isolated and wrapped by the `SourceConnectionNetworkError` custom error,
which triggers the retry logic, if enabled, in the ingest pipeline.
This PR add `include_header` argument for partition_csv and
partition_tsv. This is related to the following feature request
https://github.com/Unstructured-IO/unstructured/issues/1751.
`include_header` is already part of partition_xlsx. The work here is in
line with the current usage and testing of the `include_header` argument
in partition_xlsx.
---------
Co-authored-by: cragwolfe <crag@unstructured.io>
Fix TypeError: string indices must be integers. The `annotation_dict`
variable is conditioned to be `None` if instance type is not dict. Then
we add logic to skip the attempt if the value is `None`.
