Change how reserved columns work to have an underscore prefix

[simonw]

Right now I've implemented it such that id and commit and version and item are reserved columns, and any user-provided data with those columns gets renamed to commit_ and id_ and so-on - see #8.

I've changed my mind. I think this tool's columns should all have a _ prefix instead (like semi-private class properties in Python).

I'm even going to do that for the _id column, mainly so I don't have to explain a special case for id.

[simonw]

I'm also going to change how "banned" columns work. Currently if your data has a item_ column the tool raises an error. The new reserved column will be _item but my rule will be that if your column is reserved I'll add a trailing _ to it - and if you also have a _item_ column I'll rename that to _item__ and so-on.

Current implementation:

git-history/git_history/cli.py

Lines 288 to 300 in d58f446

	def fix_reserved_columns(item, allow_id=False, allow_banned=False):
	reserved = {"item", "version", "commit", "rowid"}
	banned = {"id_", "item_", "version_", "commit_"}
	if not allow_id:
	reserved.add("id")
	if not allow_banned and any(key in banned for key in item):
	raise click.ClickException(
	"Column {} is one of these banned columns: {}\n{}".format(
	sorted([key for key in item if key in banned]),
	sorted(banned),
	json.dumps(item, indent=4, default=str),
	)
	)

[simonw]

I'm going to split out some of this logic into a utils.py module, and write some separate tests for it.

[simonw]

Well this was a mistake... Datasette treats columns that start with an underscore as querystring parameters it should ignore!

Switching to underscore suffixes instead.

[simonw]

... I do like underscore prefixes though. I'm tempted to fix the bug I filed in Datasette instead and keep this:

• Columns beginning with an underscore do not facet correctly datasette#1506

[simonw]

I released Datasette 0.59.2 with a fix for this, so I'm free to use _id and friends here.