Allow conversion without image #26
Conversation
- add convert_file_without_image function, which can convert AWS-Textract-JSON to PAGE-XML without reading the original input image; this function takes the image’s dimensions as inputs, instead of reading these from the image - add optional “image-width” and “image-height” options to the CLI command, which triggers the use of the new function
| @click.argument("image-file", type=click.Path(dir_okay=False, exists=True)) | ||
| def cli(output_file, aws_json_file, image_file): | ||
| @click.argument("image-file", type=str) |
There was a problem hiding this comment.
Please note that with this change to the image-file argument, a path that does not exist will not be caught here. I'd welcome ideas about the best way to preserve path checking while allowing the user to supply a filename when the image is not present. Perhaps a 2nd parameter would be the best way, perhaps in the form of a dedicated command that doesn't mess with the original one? Or is this ok, and we can throw an error in the convert_file function if the referenced file can't be found? (Again, please forgive my lack of Python skills and familiarity with Python idioms.)
There was a problem hiding this comment.
Yes, interdependencies are not easily covered by Click. (It's doable via Cloup, but not necessary.)
I would rather keep the type, but merely set exists=False to skip verification of the path, and then cover all combinations in a simple assertion, see below.
| Requires the absolute dimensions of the original input image used for AWS OCR. | ||
| (If the image is available, call convert_file instead.) |
There was a problem hiding this comment.
| Requires the absolute dimensions of the original input image used for AWS OCR. | |
| (If the image is available, call convert_file instead.) | |
| Also requires the original input image used for AWS OCR, but only to reference it under | |
| `Page/@imageFilename` with its full path – does not actually require an existing file under that path. | |
| Instead, this additionally requires the absolute dimensions (pixel resolution) of the image. |
| @click.argument("image-file", type=click.Path(dir_okay=False, exists=True)) | ||
| def cli(output_file, aws_json_file, image_file): | ||
| @click.argument("image-file", type=str) |
There was a problem hiding this comment.
Yes, interdependencies are not easily covered by Click. (It's doable via Cloup, but not necessary.)
I would rather keep the type, but merely set exists=False to skip verification of the path, and then cover all combinations in a simple assertion, see below.
| @click.argument("image-file", type=str) | ||
| @click.option("--image-width", type=int, help="Width of the image in pixels, if the image isn't available.") | ||
| @click.option("--image-height", type=int, help="Height of the image in pixels, if the image isn't available.") |
There was a problem hiding this comment.
| @click.argument("image-file", type=str) | |
| @click.option("--image-width", type=int, help="Width of the image in pixels, if the image isn't available.") | |
| @click.option("--image-height", type=int, help="Height of the image in pixels, if the image isn't available.") | |
| @click.argument("image-file", type=click.Path(dir_okay=False, exists=False)) | |
| @click.option("--image-width", type=int, help="width of the image in pixels (to avoid opening the image file)") | |
| @click.option("--image-height", type=int, help="height of the image in pixels (to avoid opening the image file)") |
| The output file will reference the image file under `Page/@imageFilename` | ||
| with its full path. (So you may want to use a relative path.) | ||
| The output file will reference the image file using the name you provide. | ||
| (So you may want to use a relative path.) | ||
| """ |
There was a problem hiding this comment.
| """ | |
| """ | |
| assert image_width and image_height or image_file and exists(image_file),\ | |
| "requires passing either an existing image file path or --image-width and --image-height" |
(and also at the top of the file: from os.path import exists)
| convert_file_without_image(json_path, img_path, img_width, img_height, out_path) | ||
|
|
||
|
|
||
| def convert_file_without_image(json_path: str, img_path: str, img_width: int, img_height: int, out_path: str) -> None: |
There was a problem hiding this comment.
Note: this function should also be exported (i.e. added to the list in __init__.py's __all__)
Strange! Sounds like your |
As proposed in #25, this PR adds a
convert_file_without_imagefunction, which can convert AWS-Textract-JSON to PAGE-XML without needing to read the original input image. This function takes the image’s dimensions as inputs, instead of reading these from the image.The PR also adds “image-width” and “image-height” options to the CLI command, which triggers the use of the new function if they are supplied.
I lack python skills and cobbled this together, so pardon the lack of tests covering the new functionality.
I did run
make test-api, which passed. (Butmake test-clifailed with the same error as before my changes ("No rule to make targetOUT, needed bytest-cli. Stop.")I performed the following tests manually, confirming that the results were as expected:
Test 1: Using original features
The following command uses the original features of the utility:
... returned a 66k PAGE-XML file, with the following excerpt on line 8 referencing the input image file:
Test 2: Using the new features via CLI
The following command uses the new features added in this PR:
... returned an identical file, except for the following excerpt on line 8, corresponding to the one above:
I'd greatly appreciate any feedback!