How can I get the Group Length Elements removed during dataset writing?

[ireicht]

Similar behaviour as desribed in this issue but for function safe_as(). Although no Group Length Tags and Elements are present in the ds, when writing the DICOM file using save_as(ds), those unwanted and deprecated group-length-tags are written to the stored file.

[scaramallion]

Please attach a dataset that demonstrates the issue, or a minimally reproducible code sample.

[ireicht]

When a dicom file with GroupTagLengh is read, then the dataset hosts this tag. E.g. (0009,0000) UL 42 # 4, 1 PrivateGroupLength When I store them using save_as(), then this tag is removed.

Is there a method to enforce writing the group length tag? My script's primary task is to pseudonymize a DICOM file, and as a sanity check, the original file must be reconstructed from the pseudonymized version using the stored information about what was removed during pseudonymization.

Sorry for the mixup

[mrbean-bremen]

So you are saying that the group length tag is removed, not added? I'm confused about the actual issue.
We are indeed not writing the group length, as this has been deprecated for a long time now.

[scaramallion]

There's no method to keep the group length tags, they're removed here.

If you want to track which group length elements are removed you can do:

from pydicom.dataset import Dataset
from pydicom.dataelem import DataElement

def find_group_length_elements(ds):
    results = {}
    for elem in ds:
        if elem.tag.element == 0x0000:
            results[elem.tag] = elem
            continue

        if elem.VR == "SQ":
            for item in elem.value:
                result = find_group_length_elements(item)
                if result:
                    sq_results = results.get(elem.tag, [])
                    sq_results.append(result)
                    results[elem.tag] = sq_results

    return results


ds = Dataset()
ds[0x00090000] = DataElement(0x00090000, "UL", 4)
ds.BeamSequence = [Dataset()]
item = ds.BeamSequence[0]
item[0x00090000] = DataElement(0x00090000, "UL", 42)

for leaf, branch in find_group_length_elements(ds).items():
    print(f"{leaf}: {branch}")

(0009,0000): (0009,0000) Private Creator                     UL: 4
(300A,00B0): [{(0009,0000): (0009,0000) Private Creator                     UL: 42}]