You might want to have a look at the latest blog post

http://blog.macromates.com/2012/clever-completion/

On 19 Jan 2012, at 10:10, Hans-Jörg Bibiko wrote:

Hi,

I'm coding lots of Obj-C++ stuff. I encountered two weird behaviours.

To reproduce

• open a new window, set language to Obj-C++
• press ESC ⇢ nothing happens - fine
• press ⇥, and then ESC ⇢ "retain" will be inserted whereby I didn't type any character

furthermore 2)

• if I select something in a Obj-C++ doc and press ESC the selection will be replaced by retain

I have no idea if this is a feature or a bug but I would argue that this is a bit weird since it could happen that one presses ESC more or less accidentally. At least a selection should be untouched and ESC-completion should only be invoked if there's at least one character left from cursor as it is in TM1.

Thanks, --Hans

---

textmate mailing list textmate@lists.macromates.com http://lists.macromates.com/listinfo/textmate

On 19/01/2012, at 17.10, Hans-Jörg Bibiko wrote:

To reproduce

• open a new window, set language to Obj-C++
• press ESC ⇢ nothing happens - fine
• press ⇥, and then ESC ⇢ "retain" will be inserted whereby I didn't type any character

Since there is no prefix/suffix it inserts first completion candidate which is taken from the scoped settings (which provide a list for Objective-C wherein retain is included).

furthermore 2)

• if I select something in a Obj-C++ doc and press ESC the selection will be replaced by retain

I have no idea if this is a feature or a bug but I would argue that this is a bit weird since it could happen that one presses ESC more or less accidentally.

At least a selection should be untouched and ESC-completion should only be invoked if there's at least one character left from cursor as it is in TM1.

Esc-completion changed to allow empty prefix (since it also looks at suffix now).

I can make it a no-op when both prefix and suffix is empty to help users who accidentally press keys and expect nothing to happen — the argument seems a bit weird though ;) but given that the result from pressing ESC is likely unpredictable, I won’t claim it is a useful feature to always get a completion candidate inserted, although it could perhaps be if say, a custom completion command is setup say, for my unit test files or at root level in @implementation … @end to insert boilerplate for the only unit that makes sense given the context (i.e. method for the latter, unit test body for the former).