Updating f612cdc0..ac2a5f4c
Fast-forward
 swh/deposit/api/collection.py                    |  54 ++++++-
 swh/deposit/api/common.py                        | 187 +++++++++++------------
 swh/deposit/api/content.py                       |   4 +-
 swh/deposit/api/edit.py                          |  13 +-
 swh/deposit/api/edit_media.py                    |  15 +-
 swh/deposit/api/private/__init__.py              |  17 +--
 swh/deposit/api/private/deposit_check.py         |   7 +-
 swh/deposit/api/private/deposit_read.py          |  11 +-
 swh/deposit/api/private/deposit_update_status.py |  10 +-
 swh/deposit/api/state.py                         |   4 +-
 swh/deposit/api/sword_edit.py                    |  13 +-
 swh/deposit/tests/api/test_deposit_update.py     |   4 +-
 swh/deposit/tests/conftest.py                    |  22 +--
 13 files changed, 194 insertions(+), 167 deletions(-)

commit ac2a5f4c958d70ccf4b669fd69a441d7fec55577
Author: Valentin Lorentz <vlorentz@softwareheritage.org>
Date:   Wed Nov 25 12:59:54 2020 +0100

    conftest: Set HTTP_IN_PROGRESS to True when creating partial deposits.
    
    conftest is cheating by hard-resetting the deposit status to partial
    after creating a complete deposit; but it doesn't make sense for
    a real scenario.
    
    Additionally, a future commit will add a test relying on the
    'partial_deposit' not scheduling a task, so the HTTP_IN_PROGRESS
    header will need to be True at that point.

commit fb1c19d942ed83a2335e4a5888d6475f2d38df5b
Author: Valentin Lorentz <vlorentz@softwareheritage.org>
Date:   Wed Nov 25 11:53:27 2020 +0100

    Make APIBase.get_client return a non-Optional.
    
    As it already could never return None.

commit 42ad8df54b4e1b28dab51e934393db3a9af74862
Author: Valentin Lorentz <vlorentz@softwareheritage.org>
Date:   Wed Nov 25 11:49:59 2020 +0100

    Don't set APIBase._client in check()
    
    Instead, make it really an internal attribute with a method
    that uses as a cache.
    
    Motivation: a method named 'checks' shouldn't set attributes.

commit df87127d16c478767cd418a9d696017fdae5aed0
Author: Valentin Lorentz <vlorentz@softwareheritage.org>
Date:   Wed Nov 25 11:30:55 2020 +0100

    Remove attribute APIBase._collection
    
    Motivation:
    
    1. It was used only internally in checks() and in 'POST Col-IRI'
    2. A method named "check" should not set attributes

commit bd46805412a8d41b573aee776ac0a6b40aafa362
Author: Valentin Lorentz <vlorentz@softwareheritage.org>
Date:   Wed Nov 25 11:17:54 2020 +0100

    Check collection name when calling get_deposit_by_id, if relevant.

commit 85486a79e014f681aaab844998b1fa98ccc92b49
Author: Valentin Lorentz <vlorentz@softwareheritage.org>
Date:   Tue Nov 24 16:32:42 2020 +0100

    Move Deposit creation from api/common.py to api/collection.py
    
    Motivation:
    
    1. it's not an operation in common, only 'POST Col-IRI' may create a deposit
    2. it makes the BaseAPI._deposit_put function simpler and less complex
    
    This looks like a very long commit, but there are actually four small
    changes. In order:
    
    1. actually move that creation (the large block of code was moved
       from one file to the other)
    
    2. make the 'deposit_id' argument of most functions of 'int' type
       instead of 'Optional[int]', since we now get the id very early
       by creating it in the Col-IRI view.
    
    3. unfortunately, this means creating a Deposit and committing it
       directly to the DB, which changes the behavior of the API, as it used
       to run checks *before* committing to the DB.
       Therefore, instead of immediately committing to get the id, it passes
       the Deposit object, which will be committed later by _deposit_put if
       all checks pass.
       This means changing the 'deposit_id: int' argument of many internal
       functions to 'deposit: Deposit'. This is why the diff looks big,
       but it's actually a trivial change
    
    4. updated test_add_metadata_to_unknown_deposit because the API now returns
       the right error message, as a side-effect.