Can we try to save \n while translating text?

[CaramelHeaven]

For example, I want to translate

Lalalal

lalalal

---

Using deepl I have got -> Lalalal lalalal with cut off \n symbols and so on

[ptrstn]

Currently, this package simply passes the text as is to the deepl API. It is the deepl API itself that ignores any new line character from the input:

from deepl.api import split_into_sentences

text = """Lalalal

lalalal"""

sentences = split_into_sentences(text)
print(sentences)

Output:

['Lalalal\n\nlalalal']

from deepl.api import generate_translation_request_data

sentences = split_into_sentences(text)
data = generate_translation_request_data(
    source_language="DE", target_language="EN", sentences=sentences
)

data["params"]["jobs"][0]["raw_en_sentence"]

Output:

'Lalalal\n\nlalalal'

import json
import requests

from deepl.api import headers
from deepl.settings import API_URL

response = requests.post(API_URL, data=json.dumps(data), headers=headers)
json_response = response.json()
json_response["result"]["translations"][0]["beams"][0]["postprocessed_sentence"]

Output:

'Lalalal lalalal'  # no newline characters

I'm not quite sure yet how the web version of deepl handles the new lines in the text (I assume some javascript preprocessing). So I would have to do some text preprocessing before requesting the translation, which could lead to unexpected or corrupted behavior. This kind of code change should be thoroughly tested, which I don't have time to do at the moment.

If anyone can point me to a safe solution, I'd be happy to look into it.

[OtwakO]

I've seen something like a ignore_tag argument in the deepl documentation, is it a viable solution? Workaround way is just translating it line by line with for loop and use threading to speed up the process while keeping the line structure intact.