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/datasette/issues/2109#issuecomment-1652721111,https://api.github.com/repos/simonw/datasette/issues/2109,1652721111,IC_kwDOBm6k_c5igoXX,9599,2023-07-27T00:23:28Z,2023-07-27T00:23:37Z,OWNER,"Once again I'm tempted to formalize the extra HTML context as a `dataclass` so I can ensure it is documented correctly.

Here's an example of Hugging Face doing that: https://github.com/huggingface/transformers/blob/1689aea73346816b936b84932e12b774974e61a6/src/transformers/training_args.py#L622C1-L624

```python
from dataclasses import dataclass, field

@dataclass
class TrainingArguments:
    output_dir: str = field(
        metadata={""help"": ""The output directory where the model predictions and checkpoints will be written.""},
    )
```","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",1822934563,
https://github.com/simonw/datasette/issues/2109#issuecomment-1652715823,https://api.github.com/repos/simonw/datasette/issues/2109,1652715823,IC_kwDOBm6k_c5ignEv,9599,2023-07-27T00:17:04Z,2023-07-27T00:18:07Z,OWNER,"New decision: I had originally decided that the HTML view would just use data that could otherwise be extracted from the JSON view if you fed in enough extras.

I've changed my mind. I'm OK with the HTML view getting a few bonus things available in its context, provided those are clearly documented for template authors.

This should help avoid me having to make many changes to the templates themselves.

HTML-specific stuff will be things like `csrftoken()` and `edit_sql_url` and `show_hide_link` and `database_color` and suchlike.","{""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",1822934563,