Use integer primary keys for smaller tables

[simonw]

Refs #12. Still needs a bit more work:

• See if I can come up with a better column name than _item_hash_id (I went with _item_id)
• Ship sqlite-utils 3.19 and update dependency
• Update schema description in README
• Update reserved columns

[simonw]

This is also using sqlite-utils 3.19a0: https://github.com/simonw/sqlite-utils/releases/tag/3.19a0

git-history/setup.py

Line 34 in d645777

install_requires=["click", "GitPython", "sqlite-utils>=3.19a0"],

I need to turn that into a non-alpha release before landing this branch.

[simonw]

Here's the new schema produced by this code:

git-history/tests/test_git_history.py

Lines 194 to 219 in d645777

	assert db.schema == (
	"CREATE TABLE [commits] (\n"
	" [id] INTEGER PRIMARY KEY,\n"
	" [hash] TEXT,\n"
	" [commit_at] TEXT\n"
	");\n"
	"CREATE UNIQUE INDEX [idx_commits_hash]\n"
	" ON [commits] ([hash]);\n"
	"CREATE TABLE [items] (\n"
	" [_id] INTEGER PRIMARY KEY,\n"
	" [_item_hash_id] TEXT,\n"
	" [item_id] INTEGER,\n"
	" [name] TEXT,\n"
	" [_commit] INTEGER\n"
	");\n"
	"CREATE UNIQUE INDEX [idx_items__item_hash_id]\n"
	" ON [items] ([_item_hash_id]);\n"
	"CREATE TABLE [item_versions] (\n"
	" [_item] INTEGER REFERENCES [items]([_id]),\n"
	" [_version] INTEGER,\n"
	" [_commit] INTEGER REFERENCES [commits]([id]),\n"
	" [item_id] INTEGER,\n"
	" [name] TEXT,\n"
	" PRIMARY KEY ([_item], [_version])\n"
	");"
	)

[simonw]

The _item_hash_id column holds the hash of the ID column values for that specific item.

I called it _item_hash_id because the underlying code also calculates the hash of ALL of the column values in order to detect if it has changed since the previous version. In the Python code that's called item_hash and I was getting it confused with the other hash, hence item_hash_id as a variable name which turned into a column.

I'm sure I can come up with a better name for this column - and refactor the code once I do.

[simonw]

I should update this reserved column list:

git-history/git_history/utils.py

Line 3 in 5c06ed5

RESERVED = ("_id", "_item", "_version", "_commit", "rowid")

[simonw]

The README talks about column names and needs to be updated, but I may also extend the README to show a full schema (maybe generated with cog) for as clear an explanation as possible.

I should rig up a live demo too, ala datasette-graphql and friends.

[simonw]

I'm going to rename _item_hash_id to _item_id and in the code I'll differentiate that from _item_hash.

[simonw]

The README talks about column names and needs to be updated, but I may also extend the README to show a full schema (maybe generated with cog) for as clear an explanation as possible.

I decided not to use cog because generating an example schema requires running the full test fixtures mechanism that creates a temporary Git repository and that feels like too much complexity here.