Decreased performance in the last couple months in large modules

[jahands]

Issue Description

I don't know when exactly it started, but at some point in the last few months performance greatly went down for a large file I'm working with (>6k LOC).
Specifically, formatting speed is so slow I rarely use it anymore. Autocomplete is often quite slow that I've stopped relying on it so much as well.

I appreciate what you guys are doing! I'm happy to help in any way I can.

Attached Logs

Verbose logs:
logs.zip

Environment Information

This happens on two computers I use, one with a i7-5820k / 32GB ram, another with i7-8086k / 64GB ram. Both Windows 10 Pro

Visual Studio Code

Name	Version
Operating System	Windows_NT x64 10.0.17134
VSCode	1.26.1
PowerShell Extension Version	1.8.3

PowerShell Information

Name	Value
PSVersion	5.1.17134.165
PSEdition	Desktop
PSCompatibleVersions	1.0 2.0 3.0 4.0 5.0 5.1.17134.165
BuildVersion	10.0.17134.165
CLRVersion	4.0.30319.42000
WSManStackVersion	3.0
PSRemotingProtocolVersion	2.3
SerializationVersion	1.1.0.1

Visual Studio Code Extensions

Visual Studio Code Extensions(Click to Expand)

Extension	Author	Version
azure-account	ms-vscode	0.4.3
bash-beautify	shakram02	0.1.1
better-toml	bungcip	0.3.2
bracket-pair-colorizer	CoenraadS	1.0.59
code-beautifier	michelemelluso	2.1.0
code-settings-sync	Shan	3.0.0
csharp	ms-vscode	1.15.2
gc-excelviewer	GrapeCity	2.1.26
githistory	donjayamanne	0.4.2
Go	ms-vscode	0.6.88
hopscotch	idleberg	0.2.2
json-tools	eriklynd	1.0.2
language-liquid	neilding	0.0.3
Lua	keyring	0.0.9
material-icon-theme	PKief	3.5.3
Material-theme	zhuangtongfa	2.15.4
minify	HookyQR	0.3.0
mssql	ms-mssql	1.4.0
pdf	tomoki1207	0.3.0
PowerShell	ms-vscode	1.8.3
project-manager	alefragnani	8.0.0
python	ms-python	2018.7.1
rainbow-csv	mechatroner	0.4.2
RustyCode	saviorisdead	0.19.1
shopify-liquid-preview	kirchner-trevor	1.0.2
tcl	rashwell	0.1.0
theme-glacier	Tyriar	0.0.10
theme-happy-hipster	zadvorny	2.8.5
vsc-travis-ci-status	felixrieseberg	1.1.0
vscode-autohotkey	slevesque	0.2.2
vscode-docker	PeterJausovec	0.1.0
vscode-great-icons	emmanuelbeziat	2.1.41
vscode-liquid-snippets	killalau	1.1.1
vscode-scss	mrmlnc	0.6.2
vscode-sort-json	richie5um2	1.13.0
vscode-todo	MattiasPernhult	0.8.0
vscode-yaml-validation	djabraham	0.0.1
vsliveshare	ms-vsliveshare	0.3.577
xml	DotJoshJohnson	2.3.2

[TylerLeonhardt]

Thanks for this! As mentioned on Twitter, we've been investigating a bunch of performance related issues.

Quick question though, in addition to the large file, do you open workspaces with a ton of files?

[jahands]

Yes actually. Here's cloc output:

[rjmholt]

@jahands Does turning off CodeLens help? Let us know if it does and by how much.

[jahands]

Not it does not. I've been using Code a lot for PS the last few days and everything is still quite slow.

…

On Sat, Aug 25, 2018 at 12:55 PM Robert Holt ***@***.***> wrote: @jahands <https://github.com/jahands> Does turning off CodeLens <#1174 (comment)> help? Let us know if it does and by how much. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1497 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AKOQXUY1z2uTZiAEP_25zfPcGhRHdC55ks5uUY-JgaJpZM4WMcXn> .

[TylerLeonhardt]

@jahands just out of curiosity. Whens the last time you killed all open vscode windows and restarted it?

[jahands]

It's been a few days, probably a week on that PC actually

…

On Wed, Aug 29, 2018, 6:06 PM Tyler James Leonhardt < ***@***.***> wrote: @jahands <https://github.com/jahands> just out of curiosity. Whens the last time you killed all open vscode windows and restarted it? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1497 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AKOQXeVN-QT_qc2YSuNWy5SZwmogMBbVks5uVx5mgaJpZM4WMcXn> .

[rjmholt]

Does disabling script analysis help: "powershell.scriptAnalysis.enable": false?

[rjmholt]

Just had a look through the changes over the last few months (https://github.com/PowerShell/PowerShellEditorServices/commits/master) and there's nothing there that I can imagine causing serious performance degradations:

• Using the named pipe transport -- shouldn't change anything
• Fixed up our usage of the PSScriptAnalyzer module (by version) -- no significant extra processing there
• Changed the logging to Serilog -- used widely enough that it should be performant, and it's asynchronous anyway

[jahands]

I thought maybe it was time to split code out to multiple files so I moved a few classes to a new file. Formatting still took a really long time. But every now and again formatting will happen almost instantly like it used to. But then formatting again a little later is slow.

[rjmholt]

@jahands Have you noticed any performance improvement with the latest release?

[SeeminglyScience]

@rjmholt I think this was partially what I fixed in PowerShell/PSScriptAnalyzer#1074. Once that makes it into a release we should see a significant performance boost in a lot of scenarios.

[jahands]

@SeeminglyScience I'm happy to see that PR.

@rjmholt The format does go slightly faster, like 20 seconds instead of 50+. But it's still a long way from the <2 seconds it used to take.

[rjmholt]

@jahands See the comment above -- so far as we can tell, this particular performance regression occurred when no large changes went into PSES, so we think it was on the PSScriptAnalyzer side. @SeeminglyScience has merged a fix there, so we are now waiting on a release from them to get it downstream.

[gerane]

@TylerLeonhardt I believe I see this one. Even with codelens off. In the past, I've had the issue even with psscriptanalyzer off, but haven't tried that in a while. I will try that tonight or in the morning.

I have this problem even in relatively small modules\repos (maybe 10-20 functions) and in non workspaces and workspaces. For me, I have done tests where I've waited 10+ minutes for intellisense to finally return results. For me though, this problem has been present for well over a year. I was debugging this with @daviwil for a really long time well before he ever went to github. The best I could do was disable codelens, but even then it was still unusable.

This could potentially be different issues, but the results have always been the same for me. I really haven't been able to use intellisense for the last couple years.

I've thought it could maybe be something in my environment, but after spending months trying to remove things from my modules that I thought might cause problems, still had the same results. I was removing dynamic params, argument completers, unused modules, cleaning up profiles, etc but never did find anything that made it better.

[bergmeister]

Note that with PowerShell/PSScriptAnalyzer#1166 we expect another speed boost (twice the speed) in PSSA and we plan to release it to the wild soon (hopefully this month)

[jahands]

Hey just wanted to let let you all know that VS Code is now working really well with my large 6k+ LOC files. Autocomplete is fast, and formatting only takes 4-5 seconds (vs minutes or never finishing.) Thanks!

…

On Tue, Mar 12, 2019 at 4:16 PM Christoph Bergmeister [MVP] < ***@***.***> wrote: Note that with PowerShell/PSScriptAnalyzer#1166 <PowerShell/PSScriptAnalyzer#1166> we expect another speed boost (twice the speed) in PSSA and we plan to release it to the wild soon (hopefully this month) — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1497 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AKOQXSslhXIxDmQ8wXS2yZ7GiV6kTf7Nks5vWBk7gaJpZM4WMcXn> .

[bergmeister]

You're welcome. This is partly due to the new PSScriptAnalyzer release of 1.18.0.
@TylerLeonhardt Shall we close this issue now? Although I have accepted your invite to the this and the PSES repo, I seem to still not have contributor rights.

[rjmholt]

@bergmeister Yeah, I think we have enough other issues tracking our desire to continue improving performance. Fair to say this one tracked the degradation fixed by 1.18.0.