feat: add [figure | table] match [caption | footnote] match algorithm v2

feat: add Data api

tests/test_data/data_reader_writer/test_filebase.py

+import os
+import shutil
+
+from magic_pdf.data.data_reader_writer import (FileBasedDataReader,
+                                               FileBasedDataWriter)
+
+
+def test_filebased_reader_writer():
+
+    unitest_dir = '/tmp/magic_pdf/unittest/data/filebased_reader_writer'
+    sub_dir = os.path.join(unitest_dir, 'sub')
+    abs_fn = os.path.join(unitest_dir, 'abspath.txt')
+
+    os.makedirs(sub_dir, exist_ok=True)
+
+    writer = FileBasedDataWriter(sub_dir)
+    reader = FileBasedDataReader(sub_dir)
+
+    writer.write('test.txt', b'hello world')
+    assert reader.read('test.txt') == b'hello world'
+
+    writer.write(abs_fn, b'hello world')
+    assert reader.read(abs_fn) == b'hello world'
+    shutil.rmtree(unitest_dir)

tests/test_data/data_reader_writer/test_multi_bucket_s3.py

+import json
+import os
+
+import fitz
+import pytest
+
+from magic_pdf.data.data_reader_writer import (MultiBucketS3DataReader,
+                                               MultiBucketS3DataWriter)
+from magic_pdf.data.schemas import S3Config
+
+
+@pytest.mark.skipif(
+    os.getenv('S3_ACCESS_KEY_2', None) is None, reason='need s3 config!'
+)
+def test_multi_bucket_s3_reader_writer():
+    """test multi bucket s3 reader writer must config s3 config in the
+    environment export S3_BUCKET=xxx export S3_ACCESS_KEY=xxx export
+    S3_SECRET_KEY=xxx export S3_ENDPOINT=xxx.
+
+    export S3_BUCKET_2=xxx export S3_ACCESS_KEY_2=xxx export S3_SECRET_KEY_2=xxx export S3_ENDPOINT_2=xxx
+    """
+    bucket = os.getenv('S3_BUCKET', '')
+    ak = os.getenv('S3_ACCESS_KEY', '')
+    sk = os.getenv('S3_SECRET_KEY', '')
+    endpoint_url = os.getenv('S3_ENDPOINT', '')
+
+    bucket_2 = os.getenv('S3_BUCKET_2', '')
+    ak_2 = os.getenv('S3_ACCESS_KEY_2', '')
+    sk_2 = os.getenv('S3_SECRET_KEY_2', '')
+    endpoint_url_2 = os.getenv('S3_ENDPOINT_2', '')
+
+    s3configs = [
+        S3Config(
+            bucket_name=bucket, access_key=ak, secret_key=sk, endpoint_url=endpoint_url
+        ),
+        S3Config(
+            bucket_name=bucket_2,
+            access_key=ak_2,
+            secret_key=sk_2,
+            endpoint_url=endpoint_url_2,
+        ),
+    ]
+
+    reader = MultiBucketS3DataReader(default_bucket=bucket, s3_configs=s3configs)
+    writer = MultiBucketS3DataWriter(default_bucket=bucket, s3_configs=s3configs)
+
+    bits = reader.read('meta-index/scihub/v001/scihub/part-66210c190659-000026.jsonl')
+
+    assert bits == reader.read(
+        f's3://{bucket}/meta-index/scihub/v001/scihub/part-66210c190659-000026.jsonl'
+    )
+
+    bits = reader.read(
+        f's3://{bucket_2}/enbook-scimag/78800000/libgen.scimag78872000-78872999/10.1017/cbo9780511770425.012.pdf'
+    )
+    docs = fitz.open('pdf', bits)
+    assert len(docs) == 10
+
+    bits = reader.read(
+        'meta-index/scihub/v001/scihub/part-66210c190659-000026.jsonl?bytes=566,713'
+    )
+    assert bits == reader.read_at(
+        'meta-index/scihub/v001/scihub/part-66210c190659-000026.jsonl', 566, 713
+    )
+    assert len(json.loads(bits)) > 0
+
+    writer.write_string(
+        'unittest/data/data_reader_writer/multi_bucket_s3_data/test01.txt', 'abc'
+    )
+
+    assert 'abc'.encode() == reader.read(
+        'unittest/data/data_reader_writer/multi_bucket_s3_data/test01.txt'
+    )
+
+    writer.write(
+        'unittest/data/data_reader_writer/multi_bucket_s3_data/test02.txt',
+        '123'.encode(),
+    )
+
+    assert '123'.encode() == reader.read(
+        'unittest/data/data_reader_writer/multi_bucket_s3_data/test02.txt'
+    )

tests/test_data/data_reader_writer/test_s3.py

+import json
+import os
+
+import pytest
+
+from magic_pdf.data.data_reader_writer import S3DataReader, S3DataWriter
+
+
+@pytest.mark.skipif(
+    os.getenv('S3_ACCESS_KEY', None) is None, reason='need s3 config!'
+)
+def test_multi_bucket_s3_reader_writer():
+    """test multi bucket s3 reader writer must config s3 config in the
+    environment export S3_BUCKET=xxx export S3_ACCESS_KEY=xxx export
+    S3_SECRET_KEY=xxx export S3_ENDPOINT=xxx."""
+    bucket = os.getenv('S3_BUCKET', '')
+    ak = os.getenv('S3_ACCESS_KEY', '')
+    sk = os.getenv('S3_SECRET_KEY', '')
+    endpoint_url = os.getenv('S3_ENDPOINT', '')
+
+    reader = S3DataReader(bucket=bucket, ak=ak, sk=sk, endpoint_url=endpoint_url)
+    writer = S3DataWriter(bucket=bucket, ak=ak, sk=sk, endpoint_url=endpoint_url)
+
+    bits = reader.read('meta-index/scihub/v001/scihub/part-66210c190659-000026.jsonl')
+
+    assert bits == reader.read(
+        f's3://{bucket}/meta-index/scihub/v001/scihub/part-66210c190659-000026.jsonl'
+    )
+
+    bits = reader.read(
+        'meta-index/scihub/v001/scihub/part-66210c190659-000026.jsonl?bytes=566,713'
+    )
+    assert bits == reader.read_at(
+        'meta-index/scihub/v001/scihub/part-66210c190659-000026.jsonl', 566, 713
+    )
+    assert len(json.loads(bits)) > 0
+
+    writer.write_string(
+        'unittest/data/data_reader_writer/multi_bucket_s3_data/test01.txt', 'abc'
+    )
+
+    assert 'abc'.encode() == reader.read(
+        'unittest/data/data_reader_writer/multi_bucket_s3_data/test01.txt'
+    )
+
+    writer.write(
+        f'{bucket}/unittest/data/data_reader_writer/multi_bucket_s3_data/test02.txt',
+        '123'.encode(),
+    )
+
+    assert '123'.encode() == reader.read(
+        'unittest/data/data_reader_writer/multi_bucket_s3_data/test02.txt'
+    )

tests/test_data/io/test_s3.py

+import json
+import os
+
+import pytest
+
+from magic_pdf.data.io.s3 import S3Reader, S3Writer
+
+
+@pytest.mark.skipif(
+    os.getenv('S3_ACCESS_KEY', None) is None, reason='s3 config not found'
+)
+def test_s3_reader():
+    """test s3 reader.
+
+    must config s3 config in the environment export S3_BUCKET=xxx export S3_ACCESS_KEY=xxx export S3_SECRET_KEY=xxx
+    export S3_ENDPOINT=xxx
+    """
+
+    bucket = os.getenv('S3_BUCKET', '')
+    ak = os.getenv('S3_ACCESS_KEY', '')
+    sk = os.getenv('S3_SECRET_KEY', '')
+    endpoint_url = os.getenv('S3_ENDPOINT', '')
+    reader = S3Reader(bucket=bucket, ak=ak, sk=sk, endpoint_url=endpoint_url)
+    bits = reader.read(
+        'meta-index/scihub/v001/scihub/part-66210c190659-000026.jsonl'
+    )
+    assert len(bits) > 0
+
+    bits = reader.read_at(
+        'meta-index/scihub/v001/scihub/part-66210c190659-000026.jsonl',
+        566,
+        713,
+    )
+    assert len(json.loads(bits)) > 0
+
+
+@pytest.mark.skipif(
+    os.getenv('S3_ACCESS_KEY', None) is None, reason='s3 config not found'
+)
+def test_s3_writer():
+    """test s3 reader.
+
+    must config s3 config in the environment export S3_BUCKET=xxx export S3_ACCESS_KEY=xxx export S3_SECRET_KEY=xxx
+    export S3_ENDPOINT=xxx
+    """
+    bucket = os.getenv('S3_BUCKET', '')
+    ak = os.getenv('S3_ACCESS_KEY', '')
+    sk = os.getenv('S3_SECRET_KEY', '')
+    endpoint_url = os.getenv('S3_ENDPOINT', '')
+    writer = S3Writer(bucket=bucket, ak=ak, sk=sk, endpoint_url=endpoint_url)
+    test_fn = 'unittest/io/test.jsonl'
+    writer.write(test_fn, '123'.encode())
+    reader = S3Reader(bucket=bucket, ak=ak, sk=sk, endpoint_url=endpoint_url)
+    bits = reader.read(test_fn)
+    assert bits.decode() == '123'

tests/test_data/test_dataset.py

+
+from magic_pdf.data.dataset import ImageDataset, PymuDocDataset
+
+
+def test_pymudataset():
+    with open('tests/test_data/assets/pdfs/test_01.pdf', 'rb') as f:
+        bits = f.read()
+    datasets = PymuDocDataset(bits)
+    assert len(datasets) > 0
+    assert datasets.get_page(0).get_page_info().h > 100
+
+
+def test_imagedataset():
+    with open('tests/test_data/assets/pngs/test_01.png', 'rb') as f:
+        bits = f.read()
+    datasets = ImageDataset(bits)
+    assert len(datasets) == 1
+    assert datasets.get_page(0).get_page_info().w > 100

tests/test_data/test_read_api.py

+import os
+
+import pytest
+
+from magic_pdf.data.data_reader_writer import MultiBucketS3DataReader
+from magic_pdf.data.read_api import (read_jsonl, read_local_images,
+                                     read_local_pdfs)
+from magic_pdf.data.schemas import S3Config
+
+
+def test_read_local_pdfs():
+    datasets = read_local_pdfs('tests/test_data/assets/pdfs')
+    assert len(datasets) == 2
+    assert len(datasets[0]) > 0
+    assert len(datasets[1]) > 0
+
+    assert datasets[0].get_page(0).get_page_info().w > 0
+    assert datasets[0].get_page(0).get_page_info().h > 0
+
+
+def test_read_local_images():
+    datasets = read_local_images('tests/test_data/assets/pngs', suffixes=['png'])
+    assert len(datasets) == 2
+    assert len(datasets[0]) == 1
+    assert len(datasets[1]) == 1
+
+    assert datasets[0].get_page(0).get_page_info().w > 0
+    assert datasets[0].get_page(0).get_page_info().h > 0
+
+
+@pytest.mark.skipif(
+    os.getenv('S3_ACCESS_KEY_2', None) is None, reason='need s3 config!'
+)
+def test_read_json():
+    """test multi bucket s3 reader writer must config s3 config in the
+    environment export S3_BUCKET=xxx export S3_ACCESS_KEY=xxx export
+    S3_SECRET_KEY=xxx export S3_ENDPOINT=xxx.
+
+    export S3_BUCKET_2=xxx export S3_ACCESS_KEY_2=xxx export S3_SECRET_KEY_2=xxx export S3_ENDPOINT_2=xxx
+    """
+    bucket = os.getenv('S3_BUCKET', '')
+    ak = os.getenv('S3_ACCESS_KEY', '')
+    sk = os.getenv('S3_SECRET_KEY', '')
+    endpoint_url = os.getenv('S3_ENDPOINT', '')
+
+    bucket_2 = os.getenv('S3_BUCKET_2', '')
+    ak_2 = os.getenv('S3_ACCESS_KEY_2', '')
+    sk_2 = os.getenv('S3_SECRET_KEY_2', '')
+    endpoint_url_2 = os.getenv('S3_ENDPOINT_2', '')
+
+    s3configs = [
+        S3Config(
+            bucket_name=bucket, access_key=ak, secret_key=sk, endpoint_url=endpoint_url
+        ),
+        S3Config(
+            bucket_name=bucket_2,
+            access_key=ak_2,
+            secret_key=sk_2,
+            endpoint_url=endpoint_url_2,
+        ),
+    ]
+
+    reader = MultiBucketS3DataReader(bucket, s3configs)
+
+    datasets = read_jsonl(
+        f's3://{bucket}/meta-index/scihub/v001/scihub/part-66210c190659-000026.jsonl',
+        reader,
+    )
+    assert len(datasets) > 0
+    assert len(datasets[0]) == 10
+
+    datasets = read_jsonl('tests/test_data/assets/jsonl/test_01.jsonl', reader)
+    assert len(datasets) == 1
+    assert len(datasets[0]) == 10
+
+    datasets = read_jsonl('tests/test_data/assets/jsonl/test_02.jsonl')
+    assert len(datasets) == 1
+    assert len(datasets[0]) == 1

tests/test_model/test_magic_model.py

+import json
+
+from magic_pdf.data.read_api import read_local_pdfs
+from magic_pdf.model.magic_model import MagicModel
+
+
+def test_magic_model_image_v2():
+    datasets = read_local_pdfs('tests/test_model/assets/test_01.pdf')
+    with open('tests/test_model/assets/test_01.model.json') as f:
+        model_json = json.load(f)
+
+    magic_model = MagicModel(model_json, datasets[0])
+
+    imgs = magic_model.get_imgs_v2(0)
+    print(imgs)
+
+    tables = magic_model.get_tables_v2(0)
+    print(tables)
+
+
+def test_magic_model_table_v2():
+    datasets = read_local_pdfs('tests/test_model/assets/test_02.pdf')
+    with open('tests/test_model/assets/test_02.model.json') as f:
+        model_json = json.load(f)
+
+    magic_model = MagicModel(model_json, datasets[0])
+    tables = magic_model.get_tables_v2(5)
+    print(tables)
+
+    tables = magic_model.get_tables_v2(8)
+    print(tables)