digitalmars.D.learn - refactoring issues
- Ivan Kazmenko (24/24) Mar 17 2015 Hi,
- Vladimir Panteleev (7/10) Mar 19 2015 I did make something simple for myself, but it doesn't work in
- Ivan Kazmenko (6/16) Mar 19 2015 Hey, I also happen to use Far Manager and its internal editor, at
- Vladimir Panteleev (10/14) Mar 19 2015 It's all a bit of a mess... A FAR Lua macro[1] runs dcheck, which
- Ivan Kazmenko (11/28) Mar 20 2015 Thanks. I was able to reproduce the workflow you showed in the
- Vladimir Panteleev (4/9) Mar 20 2015 OK, here it is:
- Vladimir Panteleev (8/17) Mar 20 2015 It's used like this:
- Ivan Kazmenko (13/33) Mar 22 2015 Yay, I finally got it working! Thank you, it feels nice.
Hi, I was just refactoring a project to compile under 2.067. The fixes themselves were trivial: just adding "import std.traits;" to some files. Apparently its pieces were publicly imported by another module in 2.066. So, it's the right fix anyway. Understanding what happened, however, took more time than fixing. Surely, the compiler couldn't just say "hey, I know, you forgot to import std.traits!", and threw some distant consequences at me instead. Essentially, there were two complications: 1. In a complex template constraint, if some identifier disappears out of scope, the whole constraint silently fails, and the whole template instantiation fails with a rather generic error message. 2. When the compiler could not find a suitable overload of a function, if there are template and non-template overloads, it lists only non-template overloads. My question therefore is: can the process be improved somehow to address the above issues? For the former problem, is there a tool which jumps out and tells you use Phobos without importing things properly, or suggests a Phobos import by the name of the stuff. For the latter one, I wonder if this is the intended behavior. Ivan Kazmenko.
Mar 17 2015
On Tuesday, 17 March 2015 at 15:11:02 UTC, Ivan Kazmenko wrote:For the former problem, is there a tool which jumps out and tells you use Phobos without importing things properly, or suggests a Phobos import by the name of the stuff.I did make something simple for myself, but it doesn't work in the more complicated cases you mentioned (contracts failing because of missing imports). http://dump.thecybershadow.net/e91be687ebaeb0171d830025adf82848/autofix.gif I could post the code but the editor integration part is pretty specific to my editor.
Mar 19 2015
On Thursday, 19 March 2015 at 10:21:09 UTC, Vladimir Panteleev wrote:On Tuesday, 17 March 2015 at 15:11:02 UTC, Ivan Kazmenko wrote:Hey, I also happen to use Far Manager and its internal editor, at least for simple projects. Is that dcheck triggering a Far plugin? I have a bit of experience with Far recording macros, but didn't try to write a plugin.For the former problem, is there a tool which jumps out and tells you use Phobos without importing things properly, or suggests a Phobos import by the name of the stuff.I did make something simple for myself, but it doesn't work in the more complicated cases you mentioned (contracts failing because of missing imports). http://dump.thecybershadow.net/e91be687ebaeb0171d830025adf82848/autofix.gif I could post the code but the editor integration part is pretty specific to my editor.
Mar 19 2015
On Thursday, 19 March 2015 at 14:32:53 UTC, Ivan Kazmenko wrote:Hey, I also happen to use Far Manager and its internal editor, at least for simple projects. Is that dcheck triggering a Far plugin? I have a bit of experience with Far recording macros, but didn't try to write a plugin.It's all a bit of a mess... A FAR Lua macro[1] runs dcheck, which is just a batch file that runs DMD piped through colorout[2], which in turn parses the compiler output and records the results to a temporary file, that's read by a second program that parses and caches the .json files generated as part of my D build process and then generates editing instructions read back by the Lua macro. [1]http://dump.thecybershadow.net/5c0afb5c0f4306b1368dd8528c838133/F9.lua [2]http://blog.thecybershadow.net/2013/07/27/colorize-your-compilers-output/
Mar 19 2015
On Thursday, 19 March 2015 at 16:06:31 UTC, Vladimir Panteleev wrote:On Thursday, 19 March 2015 at 14:32:53 UTC, Ivan Kazmenko wrote:Thanks. I was able to reproduce the workflow you showed in the gif to the part where an error pop-up (e.g. "no property iota for type int") is followed by suggesting the appropriate fix. Do I need another tool for that? Colorout does not seem to suggest fixes. Also, what about this?Hey, I also happen to use Far Manager and its internal editor, at least for simple projects. Is that dcheck triggering a Far plugin? I have a bit of experience with Far recording macros, but didn't try to write a plugin.It's all a bit of a mess... A FAR Lua macro[1] runs dcheck, which is just a batch file that runs DMD piped through colorout[2], which in turn parses the compiler output and records the results to a temporary file, that's read by a second program that parses and caches the .json files generated as part of my D build process and then generates editing instructions read back by the Lua macro. [1]http://dump.thecybershadow.net/5c0afb5c0f4306b1368dd8528c838133/F9.lua [2]http://blog.thecybershadow.net/2013/07/27/colorize-your-compilers-output/2. When the compiler could not find a suitable overload of a function, if there are template and non-template overloads, it lists only non-template overloads.Since no one replied by rationalizing that it works as intended, perhaps I should file an issue about it? Ivan Kazmenko.
Mar 20 2015
On Friday, 20 March 2015 at 18:05:07 UTC, Ivan Kazmenko wrote:Thanks. I was able to reproduce the workflow you showed in the gif to the part where an error pop-up (e.g. "no property iota for type int") is followed by suggesting the appropriate fix. Do I need another tool for that? Colorout does not seem to suggest fixes.OK, here it is: https://github.com/CyberShadow/AutoFix It has some hardcoded paths though.
Mar 20 2015
On Friday, 20 March 2015 at 18:36:19 UTC, Vladimir Panteleev wrote:On Friday, 20 March 2015 at 18:05:07 UTC, Ivan Kazmenko wrote:It's used like this: C:\Path\To\colorout\colorout ^ --json C:\Temp\colorout.json ^ --trigger id=C:\Path\To\AutoFix\autofix.exe ^ C:\Path\To\colorout\d.col ^ dmd, rdmd etc...Thanks. I was able to reproduce the workflow you showed in the gif to the part where an error pop-up (e.g. "no property iota for type int") is followed by suggesting the appropriate fix. Do I need another tool for that? Colorout does not seem to suggest fixes.OK, here it is: https://github.com/CyberShadow/AutoFix It has some hardcoded paths though.
Mar 20 2015
"Ivan Kazmenko" <gassa mail.ru> On Friday, 20 March 2015 at 18:37:57 UTC, Vladimir Panteleev wrote:On Friday, 20 March 2015 at 18:36:19 UTC, Vladimir Panteleev wrote:Yay, I finally got it working! Thank you, it feels nice. Theoretically, AutoFix can be extended to handle cases similar to mine, though the details may be cumbersome. The template constraint is reported by the compiler, so the identifier to be imported is right there in the error message. One may just tokenize the string and loop over the identifiers... But that will perhaps generate too much noise in suggestions. The biggest thing to learn for me however was that I can use JSON files of Phobos and druntime, which in turn can be generated by building them. Ivan Kazmenko.On Friday, 20 March 2015 at 18:05:07 UTC, Ivan Kazmenko wrote:It's used like this: C:\Path\To\colorout\colorout ^ --json C:\Temp\colorout.json ^ --trigger id=C:\Path\To\AutoFix\autofix.exe ^ C:\Path\To\colorout\d.col ^ dmd, rdmd etc...Thanks. I was able to reproduce the workflow you showed in the gif to the part where an error pop-up (e.g. "no property iota for type int") is followed by suggesting the appropriate fix. Do I need another tool for that? Colorout does not seem to suggest fixes.OK, here it is: https://github.com/CyberShadow/AutoFix It has some hardcoded paths though.
Mar 22 2015