Bug: progress_bar resource leak — not closed on exception in save_annotated_documents and download_text_from_url

[marlonbarreto-git]

Problem

Two functions in langextract/io.py create tqdm progress bars but fail to close them when an exception occurs mid-iteration, causing terminal state corruption and resource leaks.

Location 1: save_annotated_documents (lines ~113-134)

progress_bar = progress.create_save_progress_bar(
    output_path=str(output_file), disable=not show_progress
)

with open(output_file, 'w', encoding='utf-8') as f:
    for adoc in annotated_documents:   # <-- generator can raise here
        ...
        progress_bar.update(1)

progress_bar.close()   # <-- unreachable if the for-loop raises

If annotated_documents is a generator that raises (e.g., InferenceOutputError from the LLM call), the with open(...) context manager correctly closes the file, but progress_bar.close() is never called because it sits unconditionally after the with block with no try/finally.

Location 2: download_text_from_url (lines ~295-345)

if show_progress and total_size > 0:
    progress_bar = progress.create_download_progress_bar(...)

    for chunk in response.iter_content(chunk_size=chunk_size):  # <-- network error here
        if chunk:
            chunks.append(chunk)
            progress_bar.update(len(chunk))

    progress_bar.close()   # <-- unreachable on network exception

except requests.RequestException as e:
    raise requests.RequestException(...) from e   # <-- progress_bar still open

If response.iter_content() raises a requests.RequestException (connection reset mid-download), the except block re-raises without calling progress_bar.close().

Impact

• Terminal state corruption: tqdm modifies terminal cursor state and escape sequences; without .close(), the terminal may be left in a broken state (no newline, cursor hidden, partial progress line)
• Resource leak: tqdm holds file descriptor references and I/O streams that are not released until garbage collection
• Common scenario: Any interrupted LLM extraction pipeline or flaky network download will trigger this

Proposed Fix

Wrap both in try/finally:

# save_annotated_documents
progress_bar = progress.create_save_progress_bar(...)
try:
    with open(output_file, 'w', encoding='utf-8') as f:
        for adoc in annotated_documents:
            ...
            progress_bar.update(1)
finally:
    progress_bar.close()

# download_text_from_url
progress_bar = progress.create_download_progress_bar(...)
try:
    for chunk in response.iter_content(chunk_size=chunk_size):
        if chunk:
            chunks.append(chunk)
            progress_bar.update(len(chunk))
finally:
    progress_bar.close()

Happy to submit a PR for both fixes.