[QTI] Restoring a published channel loses or breaks its QTI exercises #6211

❌ This issue is not open for contribution. Visit Contributing guidelines to learn about the contributing process and how to find suitable issues.

Overview

restore_channel rebuilds an exercise's questions only from its Perseus archive. An exercise published as a QTI package alone restores with no questions. A dual-published exercise restores with its QTI item identifiers as assessment_ids, and publishing the restored channel then fails on them.

Complexity: Medium
Target branch: unstable

Context

• Restore reads only preset='exercise' files:

  studio/contentcuration/contentcuration/utils/import_tools.py

  Lines 456 to 461 in 1beec6e

  	sql_command = (
  	"SELECT checksum, extension "
  	"preset FROM {table} WHERE contentnode_id='{id}' AND preset='exercise';".format(
  	table=FILE_TABLE, id=contentnode.node_id
  	)
  	)

• Restored assessment_ids are the ids exercise.json lists:

  studio/contentcuration/contentcuration/utils/import_tools.py

  Lines 516 to 524 in 1beec6e

  	for index, assessment_id in enumerate(data["all_assessment_items"]):
  	with open("{}.json".format(assessment_id), "rb") as fobj:
  	assessment_item = generate_assessment_item(
  	assessment_id,
  	index,
  	data["assessment_mapping"][assessment_id],
  	json.load(fobj),
  	download_url=download_url,
  	)

• Publishing decodes every legacy item's assessment_id as hex:

  studio/contentcuration/contentcuration/utils/assessment/qti/convert.py

  Lines 65 to 70 in 1beec6e

  	def hex_to_qti_id(hex_string):
  	"""
  	Encode a 32 digit hex to a 22 character base64 encoded id and a K prefix.
  	"""
  	bytes_data = bytes.fromhex(hex_string)
  	return f"K{base64.urlsafe_b64encode(bytes_data).decode('ascii').rstrip('=')}"

• On 1beec6ecb, a dual-published node with one QTI editor choice question and one legacy input_question, restored from its Perseus archive, then published:

  restored assessment_ids: item_1, <32-char hex>
  ValueError: non-hexadecimal number found in fromhex() arg at position 0
  

The Change

• Restore should recreate the questions of every published exercise, whichever archives it was published with.

How to Get There

• Publish a channel holding an exercise with one QTI editor question and one legacy question.
• Run manage.py restore_channel <channel_id> --download-url <studio url>.
• Publish the restored channel.

Acceptance Criteria

• An exercise published as a QTI package only restores with all its questions.
• An exercise mixing QTI editor and legacy questions restores with all its questions.
• Publishing a channel restored from either case succeeds.
• import_tools tests cover both cases.

AI usage

Claude Code found this during an automated self-review of the #6205 branch, reproduced the publish failure with a scratch test, and drafted this issue.