issue_comments: 655673896
This data as json
| html_url | issue_url | id | node_id | user | created_at | updated_at | author_association | body | reactions | issue | performed_via_github_app |
|---|---|---|---|---|---|---|---|---|---|---|---|
| https://github.com/simonw/sqlite-utils/issues/121#issuecomment-655673896 | https://api.github.com/repos/simonw/sqlite-utils/issues/121 | 655673896 | MDEyOklzc3VlQ29tbWVudDY1NTY3Mzg5Ng== | 9599 | 2020-07-08T18:08:11Z | 2020-07-08T18:08:11Z | OWNER | I'm with you on most of this. Completely agreed that the CLI should do everything in a transaction. The one thing I'm not keen on is forcing calling code to explicitly start a transaction, for a couple of reasons: 1. It will break all of the existing code out there 2. It doesn't match to how I most commonly use this library - as an interactive tool in a Jupyter notebook, where I'm generally working against a brand new scratch database and any errors don't actually matter So... how about this: IF you wrap your code in a `with db:` block then the `.insert()` and suchlike methods expect you to manage transactions yourself. But if you don't use the context manager they behave like they do at the moment (or maybe a bit more sensibly). That way existing code works as it does today, lazy people like me can call `.insert()` without thinking about transactions, but people writing actual production code (as opposed to Jupyter hacks) have a sensible way to take control of the transactions themselves. | {"total_count": 1, "+1": 1, "-1": 0, "laugh": 0, "hooray": 0, "confused": 0, "heart": 0, "rocket": 0, "eyes": 0} | 652961907 |