imag/libimagruby/README.md

107 lines
3.2 KiB
Markdown
Raw Normal View History

2017-01-18 18:59:18 +00:00
# imag-ruby
A Ruby gem for scripting imag modules.
2017-01-22 12:17:13 +00:00
## How does this work?
2017-01-18 18:59:18 +00:00
2017-01-22 12:17:13 +00:00
Well, as we have some problems with lifetimes here, we have a fairly complex
codebase in this crate.
2017-01-18 18:59:18 +00:00
2017-01-22 12:17:13 +00:00
### The Problem
The Problem is, that `libimagstore::store::FileLockEntry<'a>` has a lifetime. If
we would wrap this object into a ruru wrapper and pass to the Ruby code, we
couldn't guarantee anymore that the lifetime holds.
The problem is simple, you see...
### The solution?
Never pass anything to the Ruby code.
Yes, exactly. The Ruby code only sees 'handles'. It never actually gets the
`Store` object either.
We move the `Store` Object into a `Cache` object (actually, the Ruby code could
have multiple `Store` objects to work with this way) and return a `StoreHandle`
to the Ruby code (which is a UUID underneath).
Also, the Ruby code never actually touches a `FileLockEntry` - it only gets a
Handle for each `FileLockEntry` - which is a tuple of the `StoreHandle` and the
`libimagstore::storeid::StoreId` for the Entry.
Each operation on a `FileLockEntry` is then wrapped by this very library. Each
time `FileLockEntry` is touched, this library fetches the appropriate `Store`
object from the static `Cache`, then fetches the `FileLockEntry` object from it,
does the operation and then drops the object (which implies that the actual
`FileLockEntry` is `update()`d!).
### The Hell?
Yes, I know this is a lot of overhead. But what are we talking about here? This
is Ruby code we're talking about here, so speed is not our concern.
You could argue this is a hell of complexity introduced in this library and yes
it is.
If there are bugs (and I bet there are) they would be complex as hell.
But that's it... if you have a better approach, please file a PR.
## Tests?
We have tests Ruby scripts in `./test`, they are not executed by travis-ci, as
we need Ruby `2.3.0` for this and travis has `2.2.0` as latest version.
But I hope we get it in travis soonish.
## Ruby gem?
This crate will contain both the Rust bindings for imag using `ruru` and a bunch
of wrapper code for the actual `imag` gem.
We are not there yet, though.
### Why another layer of indirection?
As "ruru" does not yet support modules (which is sad btw) we would end up with
2017-01-18 18:59:18 +00:00
functions for all the things.
2017-01-22 12:17:13 +00:00
E.G.: `imag_runtime_setup()` instead of `Imag::Runtime::setup()`
2017-01-18 18:59:18 +00:00
I want to add a Ruby gem to wrap these things.
2017-01-22 12:17:13 +00:00
So basically a piece of Ruby which uses `imag.rb` (the Rust gem) to build
2017-01-18 18:59:18 +00:00
`imag` as a gem which then exports a fine module system.
2017-01-22 12:17:13 +00:00
### Ideas for module system:
2017-01-18 18:59:18 +00:00
```text
2017-01-22 12:17:13 +00:00
Imag (Module)
Runtime (Class)
Store (Class)
Entry (Class)
EntryHeader (Class)
EntryContent (Class (inherits from String))
StoreId (Class)
2017-01-18 18:59:18 +00:00
```
I would name the types the same as in the Rust codebase, to avoid confusion.
Only exception would be the `Entry` class, which would be a `FileLockEntry`
2017-01-22 12:17:13 +00:00
underneath.
If we adapt `libimagentrytag` and the other `libimagentry*`
2017-01-18 18:59:18 +00:00
libraries, we would extend this type.
2017-01-22 12:17:13 +00:00
## More plans
I want to pull these libraries into the Ruby bindings:
* libimagentryedit
* libimagentryfilter
* libimagentrylink
* libimagentrylist
* libimagentrymarkdown
* libimagentrytag
* libimagentryview
Which all provide functions on top of `libimagstore::store::{FileLock,}Entry`,
so we will implement them on `Imag::Entry`.