hsdev-0.3.0.0 by mvoidex · Pull Request #386 · SublimeHaskell/SublimeHaskell

mvoidex · 2018-01-04T04:12:46Z

@bscottm, created PR just to mention about huge updates in hsdev.

In this PR i've updated backend to use new API and to test hsdev-0.3.0.0

now uses haskell-names to resolve names in sources, and thus go to definition works for local bindings
now stores data in sqlite, so in my case user-db + package-db + project takes ~30Mb
hsdev doesn't drop ghc session anymore, so consecutive checks of same module (or modules within same project) are much faster
added new commands:

usages - get all uses of symbol (also, as long as all data is in sql, you can easily inspect it, for example i made query to find symbols, that are used least (2 - 1 for declaration and 1 for export list), and so removed many unused stuff :)
whoat - same as whois, but accepts location instead of name
rename - generates same output as autofixes (was autofix show), so can be applied with refactor command or by editor itself

Also now cabal always scanned after start of inspector

…up shown

Using TextCommand instead of WindowCommand so that it updates view each call Also use whoat if position known

bscottm · 2018-01-06T05:37:23Z

@mvoidex: Probably want to resolve the conflicts before merging.

Q: Is hsdev-0.3.0.0 backwardly compatible? Or will everyone have to upgrade to 0.3.0.0 when they take this update? I'd suggest remaining backwardly compatible for a while, if possible.

# Conflicts: # event_common.py # hsdev/backend.py # hsdev/result_parse.py

mvoidex · 2018-01-09T03:54:06Z

@bscottm, hsdev-0.3.0.0 is not backward compatible as there are many internal and some API changes.
However, I think, it's possible to support both versions, but this will require some additional work.

bscottm

There is a reason for blocking -- getting the success status of the check/lint process to determine whether other post-save actions should execute. Generally, the pause is short enough that blocking isn't really an issue. So far, no complaints.

The other reason is that the hsdev receiver (hsdev/client.py) blocks until the response function returns. That effectively prevents the response function from making more requests to hsdev (because the receiver is blocked...)

Also, avoid local functions within functions unless you truly need the closure from the outer function.

mvoidex · 2018-01-09T21:04:00Z

Why not use continuations?
Pause is definitely not short, at least in my case it takes several seconds first time and about 700-800 msecs on each consecutive check.

Also, avoid local functions within functions.
Ok

bscottm · 2018-01-13T05:17:15Z

I don't disagree with continuations, on principle. In the case of check/lint, I'm trying to remember the exact reason -- I was probably trying to track down one of @Pastafarianist's "prettify_on_save doesn't work" bugs. There are a sequence of actions that depend on check/lint success before they execute, which means adding more continuations onto the check/lint code. Just easier and slightly more maintainable to make check/lint blocking.

bscottm · 2018-01-14T06:27:11Z

2.1.18 will make check/lint async again.

…source already scanned, otherwise scan it for first time

…kage-db

moodmosaic · 2018-03-23T11:56:17Z

Is there a way to somehow hide those errors from the console? (Even if I have to temporarily comment-out a few lines of code from my local SH source.)

<hsdev stderr> �[96m2018-03-23 13:54:39 +0200�[m�[m	�[m�[91mERROR�[m�[m	�[m�[95m�[m�[m:�[m�[92mupdate/cabal/task/package-db-stack�[m�[m> �[m�[mScope leaves with exception: tool 'C:\Users\Nikos\AppData\Local\Programs\stack\x86_64-windows\ghc-8.2.2\bin\ghc-pkg.exe' failed: empty output, expecting path to user package db
<hsdev stderr> CallStack (from HasCallStack):
<hsdev stderr>   onError, called at src\System\Log\Simple\Monad.hs:104:36 in simple-log-0.9.3-AwZTZEnHmtHLEWAKWQWBHK:System.Log.Simple.Monad
<hsdev stderr> �[m�[m
<hsdev stderr> �[96m2018-03-23 13:54:39 +0200�[m�[m	�[m�[91mERROR�[m�[m	�[m�[95m�[m�[m:�[m�[92mupdate/cabal/task�[m�[m> �[m�[mScope leaves with exception: tool 'C:\Users\Nikos\AppData\Local\Programs\stack\x86_64-windows\ghc-8.2.2\bin\ghc-pkg.exe' failed: empty output, expecting path to user package db
<hsdev stderr> CallStack (from HasCallStack):
<hsdev stderr>   onError, called at src\System\Log\Simple\Monad.hs:104:36 in simple-log-0.9.3-AwZTZEnHmtHLEWAKWQWBHK:System.Log.Simple.Monad
<hsdev stderr> �[m�[m
<hsdev stderr> �[96m2018-03-23 13:54:39 +0200�[m�[m	�[m�[91mERROR�[m�[m	�[m�[95m�[m�[m:�[m�[92mupdate/cabal�[m�[m> �[m�[mScope leaves with exception: tool 'C:\Users\Nikos\AppData\Local\Programs\stack\x86_64-windows\ghc-8.2.2\bin\ghc-pkg.exe' failed: empty output, expecting path to user package db
<hsdev stderr> CallStack (from HasCallStack):
<hsdev stderr>   onError, called at src\System\Log\Simple\Monad.hs:104:36 in simple-log-0.9.3-AwZTZEnHmtHLEWAKWQWBHK:System.Log.Simple.Monad
<hsdev stderr> �[m�[m
<hsdev stderr> �[96m2018-03-23 13:54:39 +0200�[m�[m	�[m�[91mERROR�[m�[m	�[m�[95m�[m�[m:�[m�[92mupdate�[m�[m> �[m�[mtool 'C:\Users\Nikos\AppData\Local\Programs\stack\x86_64-windows\ghc-8.2.2\bin\ghc-pkg.exe' failed: empty output, expecting path to user package db�[m�[m

mvoidex · 2018-03-23T12:26:27Z

@moodmosaic there's issue mvoidex/hsdev#81, i'll try to fix it this weekend

moodmosaic · 2018-03-23T13:12:29Z

Sounds great ❤️ In general is there a way to disable all output from hsdev in SH so that it doesn't show up in the ST3 console?

moodmosaic · 2018-03-23T13:13:08Z

(Even if it can be done by modifying something in the source code, I'm fine doing that in my local SH source.)

mvoidex · 2018-03-23T17:12:54Z

Uploaded hsdev-0.3.1.3 with flag --no-color to disable colorized output

…skell into hsdev-3

moodmosaic · 2018-03-24T05:59:28Z

Anyone seen this?

With the latest commits that are merged from master into this branch (hsdev-3), hdocs aren't working, unless I go manually and revert 54259e5.

Is this still the case? I'm a bit scared to just pull the latest.

moodmosaic · 2018-03-26T08:30:52Z

OK, I pulled the latest from hsdev-3 and everything seems to work fine regarding #386 (comment).

mvoidex · 2018-03-26T17:22:58Z

@bscottm do you have any suggestions, what should be done here before merge?
Right now I have no plans to modify hsdev, so it's somehow stable now.
I'm using this branch and it seems to work for me.

It's not backward compatible now. I understand that it's not good to force users to use new hsdev version, but I'm not sure if I can find enough time to support both versions.

moodmosaic · 2018-03-26T17:39:01Z

Works fine for me too. It's fine to force users to upgrade to the same major version of hsdev, according to https://semver.org.

Perhaps, it's also a good chance to mention SemVer 2.0 on the README as well, if you guys follow it.

mvoidex · 2018-03-26T17:52:00Z

I'm trying to follow https://pvp.haskell.org/
The main difference is that according to PVP major version is X.Y part instead of X (where version is X.Y.Z).

moodmosaic · 2018-03-26T18:06:59Z

Even better, then.

mvoidex added several new methods to the hsdev backend, which were not added to the HaskellBackend base class. Sync with hsdev's current method list.

Ensure that we don't go through the on_query_completions logic when the view's source isn't Haskell. Otherwise, we potentially interfere with other completion generators. Create a default implementation function generator for backend methods that don't do anything except dispatch an empty list. Cuts down on the overall complexity of the code.

Fixed the method parameters for a lot of backend functions where hsdev changed the parameters. A few places where lines were too long, needed simplification on expressions.

Added specific diagnostic output for the build.py build system ("build_system" in component_debug setting), the aysnc worker thread ("util_worker" in component_debug). Trying to track down why the `complete` hsdev backend command freezes.

1. Add additional wrapper arguments so that different 'stack.yaml' configurations can be specified when starting hsdev. This alleviates the problem when the user's project uses 8.0.2 and needs a hsdev compiled with 8.0.2. 2. Color schemes now have comments in them. Don't load them using the JSON string loader; prefer the Sublime resource loader instead. 3. Make sure that the 'docs' scan actually does its scan, even if hsdev says that it's not supported/implemented.

juanmbellini · 2018-08-11T23:13:33Z

Any news about this?

mvoidex · 2018-08-15T15:14:17Z

@juanmbellini I was out of process for a long time and not sure if we can safely merge it

I've updated hsdev and going to upload new version compatible with current stackage LTS
There's also hsdev-0.3.2.0 branch for new hsdev version.

Check & lint works for me, but there's something wrong with autocompletion. hsdev returns correct completions, but SublimeHaskell doesn't show them. I'll try to fix it as soon as possible and update branch.

mvoidex · 2018-08-15T15:48:53Z

Completions fixed: de5610f

The backend `call` function can return boolean True or False, where a list is expected in normal circumstances. This can happen, for example, if the hsdev process is killed. In such cases, the involved functions raise a TypeError. This behaviour can be seen in the on-hover and completions handlers, upon killing the hsdev process.

mvoidex · 2018-10-09T19:04:37Z

Closing in favor of #429

mvoidex added 10 commits December 29, 2017 19:46

Basic support of hsdev 3

d7bd3e3

Changed commands symbol and module

bbc3831

symbols.py: self.context is already a list

924c995

Returned back old names for autofix function, fixed parsing corrector

e24ae6f

Use module instead of removed list_modules

aab5996

Some fixes because of different functions names

0a554a0

Also now cabal always scanned after start of inspector

Added docs and position info to UnknownSymbol

9b6e75e

Always first show_popup to set callbacks even if view has other pop…

517996b

…up shown

Fixed 'SublimeHaskellGoTo':

ae67bf3

Using TextCommand instead of WindowCommand so that it updates view each call Also use whoat if position known

Update mark positions

17f8528

mvoidex added 2 commits January 9, 2018 05:42

Merge remote-tracking branch 'origin' into hsdev-3

93b078f

# Conflicts: # event_common.py # hsdev/backend.py # hsdev/result_parse.py

Infer types on view load and after save

f1099ee

mvoidex added 6 commits January 9, 2018 07:47

Don't block on check & lint

313f3c1

Added commands to scan docs / infer types in current module

4af20cc

Fixed SublimeHaskellInferDocs command args

83d93d7

Fixed SublimeHaskellInferDocs: forgot return

3f59b7a

Added basic command to show symbol usages

faa865c

Fly check fixes

4a18f64

bscottm reviewed Jan 9, 2018

View reviewed changes

Open persistent view on going to decl via popup, like F12 does

be4fc5b

mvoidex added 5 commits January 14, 2018 18:54

on_post_save: skip scan (in case of backend with autorescan) only if …

c3c8808

…source already scanned, otherwise scan it for first time

Added command to eval expression in context of file

94ea8c6

Run infer types on hover if enable_infer_types enabled

6ee4a48

Fixed lookup call: invalid argument name used

f8b90f3

Fixed import list autocompletion: module location doesn't contain pac…

1d83cd9

…kage-db

Return result from on_query_context functions

f3311b4

mvoidex and others added 4 commits March 23, 2018 20:14

Start hsdev with --no-color flag to disable colorized output

1e8d2e0

Merge branch 'master' into hsdev-3

96a07d7

Merge branch 'hsdev-3' of https://github.com/SublimeHaskell/SublimeHa…

6967154

…skell into hsdev-3

Merge branch 'master' into hsdev-3

8627771

B. Scott Michel added 5 commits April 17, 2018 12:21

update backend methods

6cd1ee5

mvoidex added several new methods to the hsdev backend, which were not added to the HaskellBackend base class. Sync with hsdev's current method list.

pylint fixes

d8a1b73

Fixed the method parameters for a lot of backend functions where hsdev changed the parameters. A few places where lines were too long, needed simplification on expressions.

hsdev complete freezes

dfa53f4

Added specific diagnostic output for the build.py build system ("build_system" in component_debug setting), the aysnc worker thread ("util_worker" in component_debug). Trying to track down why the `complete` hsdev backend command freezes.

mvoidex mentioned this pull request Oct 9, 2018

hsdev-0.3.2.0, SublimeHaskell 2.2.0 #429

Closed

mvoidex closed this Oct 9, 2018

mvoidex mentioned this pull request Jul 28, 2019

hsdev 0.3.3.0, SublimeHaskell 2.2.0 #438

Merged

mvoidex deleted the hsdev-3 branch July 28, 2019 13:50

Conversation

mvoidex commented Jan 4, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bscottm commented Jan 6, 2018

Uh oh!

mvoidex commented Jan 9, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bscottm left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mvoidex commented Jan 9, 2018

Uh oh!

bscottm commented Jan 13, 2018

Uh oh!

bscottm commented Jan 14, 2018

Uh oh!

moodmosaic commented Mar 23, 2018

Uh oh!

mvoidex commented Mar 23, 2018

Uh oh!

moodmosaic commented Mar 23, 2018 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

moodmosaic commented Mar 23, 2018

Uh oh!

mvoidex commented Mar 23, 2018

Uh oh!

moodmosaic commented Mar 24, 2018

Uh oh!

moodmosaic commented Mar 26, 2018

Uh oh!

mvoidex commented Mar 26, 2018

Uh oh!

moodmosaic commented Mar 26, 2018

Uh oh!

mvoidex commented Mar 26, 2018

Uh oh!

moodmosaic commented Mar 26, 2018

Uh oh!

juanmbellini commented Aug 11, 2018

Uh oh!

mvoidex commented Aug 15, 2018

Uh oh!

mvoidex commented Aug 15, 2018

Uh oh!

mvoidex commented Oct 9, 2018

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

mvoidex commented Jan 4, 2018 •

edited

Loading

mvoidex commented Jan 9, 2018 •

edited

Loading

bscottm left a comment •

edited

Loading

moodmosaic commented Mar 23, 2018 •

edited

Loading