Community
Participate
Working Groups
The code completion and method navigation could be made available through shortcut keys as provided in Source Insight (www.sourceinsight.com) that could improve productivity and would be major improvement to the eclipse platform. Also syntax colouring could also be improved
F3 can be used for code navigation Ctrl+Space invikes code completion Could you provide more details of what you expect or what Source Insight provides? Please file a separate bug report for syntax coloring with details what to improve. Do not forget the build id.
Well to explain what source insight provides, the appearance is more or less similar to the Java perspective of eclipse. In source insight, we have the method names listed in the left pane and the source files listed in the right pane and there is a pane at the bottom which is called a context window, which shows the method defintion or the variable declaration for the method under the cursor in the editor window. The code completion is exceptionally good that i dont need to press Ctrl+Space to give me a list to choose, instead it shows the appropriate code completion list as i start typing. For eg if i have variables string1, sprite1, simpleVar and if i type s it shows me a list of all the three and any others that match within the project and if i am to declare a new variable say for a Graphics object, and if i type G it show all the classes in G and then continue typing r it shows classes that begin with Gr and so on. and i can either continue typing or choose from the list. As far as method navigation, it displays a list of methods in the left pane, in the order defined in the code pane, u also have the option to sort them. If i need to navigate to method that is way up in the top or way down in the bottom. All that i have to do is press a shortcut key Alt+L that takes me to the textbox of the list in the left pane and as i start typing the method name that i want to get to, i shortens the list and provides me with the methods that have the same characters typed and i could use the down arrow select the method and then press enter it takes me to the method. this way i dont even need to use the mouse to traverse to the wanted method. I guess this would, improve productivity a lot as most programmers are used to the keyboard for they routine tasks than the mouse, and they have to use the keyboard :). And on the right pane it lists the files in the project, the same that applies to method navigation applies to this, but the shortcut key is Ctrl+O. One more thing abt finding text is the Shift+F3 and Shift+F4 where these shortcut keys start finding the text under the cursor/selection, forward and backward respectively. Source insight should be given credit for so much of wonderful features that it has, esp version 3.5. If such features could be brought into Eclipse, it would really make eclipse a great IDE to use. I still recommend, ppl who ask for a better IDE to use Eclipse if they want a free and good IDE. Good work!! eclipse team keep it up!!! (In reply to comment #1) > F3 can be used for code navigation > Ctrl+Space invikes code completion > > Could you provide more details of what you expect or what Source Insight provides? > > Please file a separate bug report for syntax coloring with details what to improve. > > Do not forget the build id.
Could you ask some screen shots? Auto-activation for code assist is available when '.' is pressed. Offering it at any place could be interesting to investigate further.
Created attachment 23985 [details] auto complete feature of Source Insight The picture illustrates the auto complete feature of source insight, where i just start typing the first few letters and it provides me with the suitable list of class names, this also applies to variables, methods in project - have attached them separately
Created attachment 23986 [details] Method/Variable Completion List This image illustrates the variable/method name completion, see as i type the first letter it starts providing me the available list
Created attachment 23987 [details] Method Navigation using the Alt+L key This image illustrates the method navigation using the Alt+L (shortcut key - Configurable) where in the text box i typed a few characters and source insight shortlists the methods and i can use the down arrow to select the appropriate method and the press enter and it will take me to the method i chose in the editor screen (in the center)
Sure, have attached 3 images for that, i hope that should explain almost everything about Source Insight, the main editor window, the method list, the project files and the context window in the bottom of the screen, that shows the appropriate context. I would suggest that u could try a trial version that has all the full functionality enabled and get the feel of it. Thanks, R.Prabhu
Did you try Ctrl+O (also configurable)? This gives you all the methods, and you can also type to narrow down the search. If you type Ctrl+O a second time you'll even get all inherited members and there are various ways to sort the list. Similar, if you select a member or type in the editor press Ctrl+F3. In both quick views you can select to navigate to that member.
Are u asking it with reference to Eclipse or Source Insight. As far as Source insight is concerned Ctrl+O opens the files in Project and indeed all the shortcut keys are configurable. If in case u mean in it in eclipse, i will try that too, thanks
I meant to try Ctrl+O and Ctr+F3 in Eclipse. I think you'll love it ;-)
Thanks Daniel, i tried it and indeed its a cool feature ;-), of showing inherited members when Ctrl+O being pressed again, that i havent found in other IDEs. But, would love to have the auto code-completion everywhere in the Editor as in Source Insight. (if possible in all the plugins, esp CDT). Also is there a way to search the word under the cursor using a shortcut key, without selecting it. Though we have that for Ctrl+K and Ctrl+Shift+K for selected work to search Next occurence and previous occurence. Thanks (In reply to comment #10) > I meant to try Ctrl+O and Ctr+F3 in Eclipse. I think you'll love it ;-)
>Also is there a >way to search the word under the cursor using a shortcut key, without selecting >it. Though we have that for Ctrl+K and Ctrl+Shift+K for selected work to search >Next occurence and previous occurence. Thanks No this is currently not possible in general. In the Java Editor there's Ctrl+Shift+U which searchers for occurrences. Also in the Java editor, you can enable occurrence marking: use the marker in the tool bar to enable and disable it and the Java > Editor > Mark Occurrences preferences to control it.
> No this is currently not possible in general. In the Java Editor there's > Ctrl+Shift+U which searchers for occurrences. Also in the Java editor, you can > enable occurrence marking: use the marker in the tool bar to enable and disable > it and the Java > Editor > Mark Occurrences preferences to control it. Yeah, i saw that too!! Marking is indeed a wonderful feature rather i've not found that in any other IDE that i've used. Thanks Daniel.
Hi Daniel, Is there any plan to implement the content-assist-everywhere in Eclipse 3.2 Release or has it been scheduled later? Also it would be better if it is available for all the common programming languages like C++, Java, etc Thanks, R.Prabhu
Once you get the list, what happens if you continue typing and - if what you type is in the list? - if what you type is not in the list? Is the list moved to the right while typing?
Comments Inlined For explanation purpose we shall have the following variables val1, val2, val3, var1, var2, vlx1, vlx2 (variable type does not matter) (In reply to comment #15) > Once you get the list, what happens if you continue typing and > - if what you type is in the list? when i type v, it displays a list with all the variables that start with v i.e., val1, val2, val3, var1, var2, vlx1, vlx2 when i type 'a' after that, all the variable that start with 'va' are displayed in that list, and the others (vlx1, vlx2) are removed. In short only those variables that form a superset of the pattern are displayed in the list > - if what you type is not in the list? When i type something not in the list, in the foregoing example, if i type 'e' after 'v', it list is populated with the class and interface names that start with 'v' (case-insensitive), but when the name/variable is inserted it is inserted with proper casing. For example in this case when i type 've' it shows me a list with "Vector, VectorEnumerator, VerifyError, Version" etc > Is the list moved to the right while typing? Yes the list is moved to the right while typing I hope this info is helpful
There are no plans to add this for 3.2.
I understand the complexity. Hope that i will be added later :) Thanks Daniel (In reply to comment #17) > There are no plans to add this for 3.2.
I mean it will be added later :) (In reply to comment #18) > I understand the complexity. Hope that i will be added later :) Thanks Daniel > (In reply to comment #17) > > There are no plans to add this for 3.2.
See Bug #169954 for a plugin that implements this feature by extending auto-activation to all alpha characters.
Sorry. See Bug #159157.
*** Bug 226603 has been marked as a duplicate of this bug. ***
*** Bug 250784 has been marked as a duplicate of this bug. ***
I've solved this problem by moving to KDevelop which has instant auto-completion and works really fast... :-P
This is #1 on my feature request list as well. The Code Assist of java and PHP both supports this option natively in Eclipse, so the platform CA performance is definitely not an issue. I have just tried it in java and it works fluently. It even nicely pops up help on system functions as I type. Considering the fact that now practically *all* major competing IDEs support this feature, probably this is the #1 objection against using Eclipse+CDT at the moment for C++ development.
(In reply to comment #25) > Considering the fact that now practically *all* major competing IDEs support > this feature, probably this is the #1 objection against using Eclipse+CDT at > the moment for C++ development. For the CDT-specific case, I think you want bug 310372.
*** Bug 358031 has been marked as a duplicate of this bug. ***
#cc remove
*** Bug 534648 has been marked as a duplicate of this bug. ***
*** Bug 536991 has been marked as a duplicate of this bug. ***
I don't think this belongs in Platform Text. Finding when a content-assist session can be automatically triggered usefully really depends on the language. For example, in Java, one wouldn't want a completion startup after a " or a { because their is either no result, or context is to vague to print something useful. Similarly, in XML, one wouldn't really want a completion on > because there isn't much to show in the vast majority of cases. With these examples, this shows that it's a language specific behavior we want, and that it's already covered properly in Platform Text by giving the ability to specific editors/content-assist processors to define trigger chars. Looking at the initial comments, I'm retargettting this to JDT specifically, while similar bugs can be opened for some other editors. I think it's actually a case for bug 508789 and related bugs containing concrete steps for auto-activation of content-assist. We're close to have a good auto-activation strategy in JDT, that could easily become the default once several patches are addressed.
*** Bug 547740 has been marked as a duplicate of this bug. ***
I don't think bug 531061 is a blocker for this one.
(In reply to Mickael Istria from comment #33) > I think it's actually a case for bug 508789 and related bugs containing > concrete steps for auto-activation of content-assist. I agree we should start with bug 508789.
(In reply to Dani Megert from comment #35) > I don't think bug 531061 is a blocker for this one. Without bug 531061, content-assist everywhere is almost not usable on most machines as it freezes the IDE for some time on every keystroke. On good machines/OSs, it can be only a few milliseconds and is acceptable; on slower machine, it can be up to several seconds and is totally not usable when simply typing text.
(In reply to Mickael Istria from comment #37) > (In reply to Dani Megert from comment #35) > > I don't think bug 531061 is a blocker for this one. > > Without bug 531061, content-assist everywhere is almost not usable on most > machines as it freezes the IDE for some time on every keystroke. On good > machines/OSs, it can be only a few milliseconds and is acceptable; on slower > machine, it can be up to several seconds and is totally not usable when > simply typing text. Yes, that's why it should not be the default at the beginning.
(In reply to Dani Megert from comment #38) > (In reply to Mickael Istria from comment #37) > > (In reply to Dani Megert from comment #35) > > > I don't think bug 531061 is a blocker for this one. > > > > Without bug 531061, content-assist everywhere is almost not usable on most > > machines as it freezes the IDE for some time on every keystroke. On good > > machines/OSs, it can be only a few milliseconds and is acceptable; on slower > > machine, it can be up to several seconds and is totally not usable when > > simply typing text. > Yes, that's why it should not be the default at the beginning. i.e. go step by step and start with bug 508789.
(In reply to Dani Megert from comment #38) > Yes, that's why it should not be the default at the beginning. I think we have different scopes in mind for this bug. I have in mind "enabling auto-activation by default" why it seems like you have in mind "provide way for auto-activation". Can you please refine the title a bit ?
(In reply to Mickael Istria from comment #40) > (In reply to Dani Megert from comment #38) > > Yes, that's why it should not be the default at the beginning. > > I think we have different scopes in mind for this bug. I have in mind > "enabling auto-activation by default" why it seems like you have in mind > "provide way for auto-activation". Can you please refine the title a bit ? I think we're on the same page. This one will just come after offer the option to enable it and make it stable to make that one the default.
(In reply to Dani Megert from comment #41) > I think we're on the same page. This one will just come after offer the > option to enable it and make it stable to make that one the default. As long as bug 531061 is not fixed, the default behavior should not change. If we enable content-assit by default without bug 531061 (and without checking that all content-assist processors can run asynchronously for the given editor), we'll just ship a Java editor that freezes on every keystroke, that would be worse than the current state. I'm all in favor of the option to enable; but making it on by default really requires bug 531061 to be fixed to be usable.
(In reply to Mickael Istria from comment #42) > As long as bug 531061 is not fixed, the default behavior should not change. Do we know how many people actually have slow computers and would be affected? I know that users use the hack to enable it for years and back then computers were slow. I also would not want to enable it by default if it really hits most users.
I have a fairly good machine and I'm seing freeze about every time I type a dot or a parenthesis. As a result, I've excluded it from the auto-activation list, but it's a pity because it's exactly just at those times completion is the most useful and shines the most if it's smart enough
Please ignore my ignorance, but I seem to recall discussion about what actually "activation" or "trigger chars" should mean: is it about requesting proposals or about selecting a proposal or both? Reason for asking: I sometimes find myself in situations where completion inserts proposals that I didn't ask for. Normally this happens (a) in the middle of existing program text and (b) when my fingers are typing-correcting-typing faster than I can control them, so I effectively don't know which exact sequence of keystrokes I applied :/ While proposing more eagerly is perhaps mostly a matter of performance, selecting and inserting without explicit selection is problematic and may have the opposite effect of the intended convenience. In combination: any risk of premature selection of proposals obviously weighs heavier if proposals are presented more eagerly. Perhaps, what I'm looking for is Disable insertion triggers except 'Enter' Help rightly suggests to set this to 'On' when many trigger chars are selected. BTW, the help for this options speaks of "pre-defined trigger characters" without giving a hint at which these are. I'm supporting the notion that for any behavior that's not the default today, we'd need feedback from field use by many users, which means: a simple means for opt-in as a prerequisite for this round of feedback.
(In reply to Stephan Herrmann from comment #45) > Perhaps, what I'm looking for is > Disable insertion triggers except 'Enter' Yes, this option was implemented exactly for this reason. On my IDE, I've it always turned on (insertion triggers disabled). The other way round drives me crazy for the same reason as what you described. > Help rightly suggests to set this to 'On' when many trigger chars are > selected. That seems like a good idea. Or maybe just turning this on by default. At some point, it might also be interesting in the generic content assistant framework to keep track of what is the origin of a content assist session (explicit invocation by user, auto-activation on any keystroke, auto-activation by a given trigger...) and to enable auto-commit characters or not according to the source event. It's how things are somehow designed in the Language Server Protocol for instance. But pragmatically, turning "Disable insertion triggers except 'Enter'" to On by default should be good and simple enough.
(In reply to Mickael Istria from comment #46) > (In reply to Stephan Herrmann from comment #45) > > Perhaps, what I'm looking for is > > Disable insertion triggers except 'Enter' > > Yes, this option was implemented exactly for this reason. Where do you see that option? One option that might hit you is "Insert single proposals automatically".
(In reply to Mickael Istria from comment #44) > I have a fairly good machine and I'm seing freeze about every time I type a > dot or a parenthesis. As a result, I've excluded it from the auto-activation > list, but it's a pity because it's exactly just at those times completion is > the most useful and shines the most if it's smart enough OK, I've added bug 531061 back.
(In reply to Dani Megert from comment #48) > (In reply to Mickael Istria from comment #44) > OK, I've added bug 531061 back. +2, as discussed yesterday in the PMC call, I have supported customers projects there every code completion blocks the UI for 5-15 seconds. We recently increased the scope in PDE for quick fixes via Bug 427971 and had to revert via Bug 547217 as the increased search scope created UI freezes. To quote a customer from me: "we use IntelliJ for new projects as our developer productivity is unacceptable with Eclipse for any real sized projects". The above customer had hopes in the new "Google index" which unfortunately never worked.
> (In reply to Mickael Istria from comment #44) > OK, I've added bug 531061 back. Thank you. I have changed the settings to auto-activate with every printable keystroke. I think I even hacked the code to support this. Before I upgraded my laptop 6 months ago, my laptop could handle the auto-activate very well. Yes, sometimes there was a UI pause with each keystroke and it was annoying. My newer laptop has the same type of processor yet newer. Intel reduced the power usage and hence the performance of the processor. I am regretting the upgrade. Next time I upgrade I will probably spend my own money to get a really good processor instead of the standard issue from my company. With this newer processor, the UI pause happens a lot more. I am getting annoyed but I haven't turned it off yet. The feature is extremely useful. I can type a 2-3 characters and then hit enter to select the top choice. This makes programming much faster. On the other hand, sometimes I don't need autocomplete and simply type out code. I switch back and forth between using autocomplete and typing quickly. It depends on which I can do faster. Please make autocomplete asynchronous from the typing. This way I can type at full speed and not be hindered by the UI pauses. Please make autocomplete very fast. Ideally, it would be < 125 ms which is my typing speed at 80 words per minute (WPM). (At 300 WPM, the speed of very fast typists, autocomplete would need to be 33 ms.) If it were that fast, then I could incorporate autocomplete into my typing and not miss a beat.
(In reply to Nathan Reynolds from comment #50) Nathan, thanks for your comment.
(In reply to Dani Megert from comment #47) > (In reply to Mickael Istria from comment #46) > > (In reply to Stephan Herrmann from comment #45) > > > Perhaps, what I'm looking for is > > > Disable insertion triggers except 'Enter' > > > > Yes, this option was implemented exactly for this reason. > Where do you see that option? Preferences > Java > Editor > Content Assist > Insertion Thinking more about it there is *one* trigger for insertion which I like (in addition to 'Enter'): ';'
(In reply to Stephan Herrmann from comment #52) > (In reply to Dani Megert from comment #47) > > (In reply to Mickael Istria from comment #46) > > > (In reply to Stephan Herrmann from comment #45) > > > > Perhaps, what I'm looking for is > > > > Disable insertion triggers except 'Enter' > > > > > > Yes, this option was implemented exactly for this reason. > > Where do you see that option? > > Preferences > Java > Editor > Content Assist > Insertion > > Thinking more about it there is *one* trigger for insertion which I like (in > addition to 'Enter'): ';' Ah, right!
*** Bug 552159 has been marked as a duplicate of this bug. ***
is there any progress on this issue since last year? or is it forgotten?
(In reply to hossein mahdavipour from comment #55) > is there any progress on this issue since last year? Not that I am aware of. > or is it forgotten? It's not forgotten, it's just not the highest priority of any active contributor at the moment. Would you be interested in trying to write a patch for this?
*** Bug 558967 has been marked as a duplicate of this bug. ***
Bug 569668 would be nice to fix before we switch to async completion by default. However this is not a full requirement as just tweaking preferences allows to dodge the issue. I believe that continuous completion more or less mandate to disable insertion trigger characters if we want to avoid some annoyance.
(In reply to Mickael Istria from comment #58) > Bug 569668 would be nice to fix before we switch to async completion by > default. However this is not a full requirement as just tweaking preferences > allows to dodge the issue. I believe that continuous completion more or less > mandate to disable insertion trigger characters if we want to avoid some > annoyance. Let me repeat here from bug 569668: that other bug is not specific to particular triggers. Even selecting a proposal with Enter and quickly continuing to type may produce complete nonsense in the buffer. It seems there is already more concurrency going on than any committer is able to understand. In my experience, in some 18 years of using Eclipse, bug 569668 is the worst that ever happened in this area. Moving forward without understanding that bug sounds like a VERY bad idea to me.
(In reply to Stephan Herrmann from comment #59) > Let me repeat here from bug 569668: that other bug is not specific to > particular triggers. Even selecting a proposal with Enter and quickly > continuing to type may produce complete nonsense in the buffer. It seems > there is already more concurrency going on than any committer is able to > understand. In my experience, in some 18 years of using Eclipse, bug 569668 > is the worst that ever happened in this area. Moving forward without > understanding that bug sounds like a VERY bad idea to me. I've to second this. It's very unfortunate that input events are no reliably applied anymore. For me, this happens regularly with content proposals and quickfixes (add import).
I added the "depends on" link. Please keep discussions about bug 569668 on its own tickets, ideally adding there some steps to reproduce or any other technical info that can help in fixing it.
(In reply to Mickael Istria from comment #61) > I added the "depends on" link. thanks > Please keep discussions about bug 569668 on its own tickets, ideally adding > there some steps to reproduce or any other technical info that can help in > fixing it. it's all in that bug already. I have no idea what additional information you could possibly be asking about.
