 |
lua-l archive |
|
|
||
|
On 4-Aug-04, at 1:52 PM, Brian Hook wrote:
Some areas are far too ambitious to roll into a standard distributions: graphical interfaces, a math library, a native threads library, high quality audio, file/directory handling, signal handling, etc.
All the same, thinking about what a directory interface would look like is a useful exercise.
For example, could directories simply be modeled as Lua tables? Or would it be too foreign for the table-set operation to throw an error?
I realise that we are probably not going to get to the level of defining a standard interface for the FreeBSD kqueue call or Mac OS X equivalent which you can use to receive notification when a new file has been added to a directory. But wouldn't it be useful to have a defined interface for common things like verifying whether a string represented a file or a directory; creating new files within directories; iterating over files within directories; and safely walking a directory path?
That said, there are certain concepts that are common across a huge range of platforms that could easily be encapsulated in a single file and would likely benefit a lot of users. The obvious ones are a ms accurate timer (when available), sleep, and maybe a handful of others.
The obvious *interface* for (more) accurate timers would be to allow os.time() to return a non-integer and insist that it be the number of seconds from some (undefined) starting time. The implementation details on some platforms, as Ben Sunshine-Hill rightly points out, might be more difficult.
Alex Sandro Queiroz e Silva asks about whether that was the goal of the LuaCheia project. I'm not part of the project so I cannot answer for them, but my understanding was that LuaCheia's goal is to provide an out-of-the-box Lua with a lot of features; that is, an implementation.
What I am suggesting here is some thinking about *interfaces*: what does a good API look like? And, furthermore, that rather than try to define a platform-independent monster (viz. various GUI libraries), just trying to focus on facilities which are very common across platforms so as to provide a common "style" for extensions.
R.