diff options
| -rw-r--r-- | bug/sorting.go | 4 | ||||
| -rw-r--r-- | doc/model.md | 72 |
2 files changed, 54 insertions, 22 deletions
diff --git a/bug/sorting.go b/bug/sorting.go index 231055bd..360a969d 100644 --- a/bug/sorting.go +++ b/bug/sorting.go @@ -18,7 +18,7 @@ func (b BugsByCreationTime) Less(i, j int) bool { // When the logical clocks are identical, that means we had a concurrent // edition. In this case we rely on the timestamp. While the timestamp might // be incorrect due to a badly set clock, the drift in sorting is bounded - // by the first sorting using the logical clock. That means that if user + // by the first sorting using the logical clock. That means that if users // synchronize their bugs regularly, the timestamp will rarely be used, and // should still provide a kinda accurate sorting when needed. return b[i].firstOp().Time().Before(b[j].firstOp().Time()) @@ -46,7 +46,7 @@ func (b BugsByEditTime) Less(i, j int) bool { // When the logical clocks are identical, that means we had a concurrent // edition. In this case we rely on the timestamp. While the timestamp might // be incorrect due to a badly set clock, the drift in sorting is bounded - // by the first sorting using the logical clock. That means that if user + // by the first sorting using the logical clock. That means that if users // synchronize their bugs regularly, the timestamp will rarely be used, and // should still provide a kinda accurate sorting when needed. return b[i].lastOp().Time().Before(b[j].lastOp().Time()) diff --git a/doc/model.md b/doc/model.md index 9f28c4dd..6f171c5b 100644 --- a/doc/model.md +++ b/doc/model.md @@ -11,6 +11,11 @@ Instead of storing directly the final bug data, we store a series of edit `Opera ```json { "type": "SET_TITLE", + "author": { + "name": "René Descartes", + "email": "rene@descartes.fr" + }, + "timestamp": 1533640589, "title": "This title is better" } ``` @@ -34,35 +39,62 @@ Here is the complete picture: | | | - +-----------+ +-----------+ "ops" +-----------+ - | Commit |----------> Tree |-------|------------| Blob | (OperationPack) - +-----------+ +-----------+ | +-----------+ - | | - | | - | | "root" +-----------+ - +-----------+ +-----------+ |------------| Blob | (OperationPack) - | Commit |----------> Tree | | +-----------+ - +-----------+ +-----------+ | - | | - | | "media" +-----------+ +-----------+ - | +------------| Tree |--->| Blob | bug.jpg - +-----------+ +-----------+ +-----------+ +-----------+ - | Commit |----------> Tree | - +-----------+ +-----------+ + +-----------+ +-----------+ "ops" +-----------+ + | Commit |----------> Tree |---------+------------| Blob | (OperationPack) + +-----------+ +-----------+ | +-----------+ + | | + | | + | | "root" +-----------+ + +-----------+ +-----------+ +------------| Blob | (OperationPack) + | Commit |----------> Tree |-- ... | +-----------+ + +-----------+ +-----------+ | + | | + | | "media" +-----------+ +-----------+ + | +------------| Tree |---+--->| Blob | bug.jpg + +-----------+ +-----------+ +-----------+ | +-----------+ + | Commit |----------> Tree |-- ... | + +-----------+ +-----------+ | +-----------+ + +--->| Blob | demo.mp4 + +-----------+ ``` Now that we have this, we can easily merge our bugs without conflict. When pulling bug's update from a remote, we will simply add our new operations (that is, new `Commit`), if any, at the end of the chain. In git terms, it's just a `rebase`. ## You can't have a simple consecutive index for your bugs -TODO: complete when stable in the code +The same way git can't have a simple counter as identifier for it's commit as SVN do, we can't have consecutive identifiers for bugs. ---> essentially a semi-random ID + truncation for human consumption +`git-bug` use as identifier the hash of the first commit in the chain of commit of the bug. As this hash is ultimately computed with the content of the `CREATE` operation that include title, message and a timestamp, it will be unique and prevent collision. + +The same way as git does, this hash is displayed truncated to a 7 characters string to human user. Note that when specifying a bug id in a command, you can enter as few character as you want as long as there is no ambiguity. If multiple bugs match your prefix, `git-bug` will complain and display the potential matches. ## You can't rely on the time provided by other people (their clock might by off) for anything other than just display -TODO: complete when stable in the code +When in the context of a single bug, events are already ordered without the need of a timestamp. An `OperationPack` is an ordered array of operations. A chain of commit orders `OperationPack` with each other. + +Now, to be able to order bugs by creation or last edition time, `git-bug` use a [Lamport logical clock](https://en.wikipedia.org/wiki/Lamport_timestamps). A Lamport clock is a simple counter of event. When a new bug is created, its creation time will be the highest time value we are aware of plus one. This declare a causality in the event and allow to order bugs. + +When bugs are push/pull to a git remote, it might happen that bugs get the same logical time. This means that they were created or edited concurrently. In this case, `git-bug` will use the timestamp as a second layer of sorting. While the timestamp might be incorrect due to a badly set clock, the drift in sorting is bounded by the first sorting using the logical clock. That means that if users synchronize their bugs regularly, the timestamp will rarely be used, and should still provide a kinda accurate sorting when needed. ---> inside a bug, we have a de facto ordering with the chain of commit +These clocks are stored in the chain of commit of each bug, as entries in each main git `Tree`. The first commit will have both a creation time and edit time clock, while a later commit will only have an edit time clock. A naive way could be to serialize the clock in a git `Blob` and reference it in the `Tree` as `"create-clock"` for example. The problem is that it would generate a lot of blobs that would need to be exchanged later for what is basically just a number. ---> to order bugs, we can use a Lamport clock + timestamp when concurrent editing +Instead, the clock value is serialized directly in the `Tree` entry name (for example: `"create-clock-4"`). As a Tree entry need to reference something, we reference the git `Blob` with an empty content. As all of these entries will reference the same `Blob`, no network transfer is needed as long as you already have any bug in your repository. + + +Example of Tree of the first commit of a bug: +``` +100644 blob e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 create-clock-14 +100644 blob e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 edit-clock-137 +100644 blob a020a85baa788e12699a4d83dd735578f0d78c75 ops +100644 blob a020a85baa788e12699a4d83dd735578f0d78c75 root +``` +Note that both `"ops"` and `"root"` entry reference the same OperationPack as it's the first commit in the chain. + + +Example of Tree of a later commit of a bug: +``` +100644 blob e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 edit-clock-154 +100644 blob 68383346c1a9503f28eec888efd300e9fc179ca0 ops +100644 blob a020a85baa788e12699a4d83dd735578f0d78c75 root +``` +Note that the `"root"` entry still reference the same root OperationPack. Also, all the clocks reference the same empty `Blob`.
\ No newline at end of file |