STAN-0206 doesn't go away after enabling -XStrictData

[codygman]

Posting this for someone else, I'll try to add more detail (or get them too) and reproduce if I can later.

[goolord]

minimal example https://github.com/goolord/stan-stricdata-minimal

nothing special required to reproduce this really, the error doesn't go away whether it's a language pragma, in the default-extensions stanza of the cabal file, or otherwise

[chshersh]

@codygman @goolord Can you provide more info on Stan version you are using, how did you build the project, created the HIE files and how are you running Stan?

I just tried running the following on the example repository using the latest version of Stan from master:

cabal build
stan

And everything works as expected, i.e. I don't see any observations in the terminal output.

[goolord]

\Stan v0.0.1.0
 ➤ Git revision: 04d5a192cebb8e63b8704bdd2a8afe5b2d1c3493
 ➤ Commit date:  Tue Sep 15 16:11:42 2020 +0000
 ➤ GHC version:  8.8

built w/ cabal from master

[vrom911]

@codygman @goolord This is not reproducible for me as well. I build the project with Cabal and the same Stan and GHC version. I am on Mac, and @chshersh checked on Linux.

Stan does not produce any observations, and also is able to recognize the StrictData extension.

           Stan's Summary:
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┓
┃ Analysed modules             ┃ 1      ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Analysed Lines of Code       ┃ 11     ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Total Haskell2010 extensions ┃ 1      ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Total SafeHaskell extensions ┃ 0      ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Total checked inspections    ┃ 44     ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Total found observations     ┃ 0      ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Total ignored observations   ┃ 0      ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Project health               ┃ 100%   ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┻━━━━━━━━┛

Could you provide some details of any specific tools or features you are using to build/run stan-stricdata-minimal?

[vrom911]

Moreover, we have a lot of tests on this particular inspection, so we do check that this inspection actually works:

stan/test/Test/Stan/Analysis/AntiPattern.hs

Lines 77 to 137 in 04d5a19

	strictFieldsSpec :: Analysis -> Spec
	strictFieldsSpec analysis = describe "STAN-0206: Strict data type fields" $ do
	describe "Without extensions" $ do
	let checkObservation = observationAssert ["AntiPattern", "Stan0206"] analysis
	let noObservation = noObservationAssert ["AntiPattern", "Stan0206"] analysis
	it "Doesn't trigger on strict field" $
	noObservation AntiPattern.stan0206 7
	it "Finds simple lazy field" $
	checkObservation AntiPattern.stan0206 8 7 25
	it "Finds polymorphic lazy field" $
	checkObservation AntiPattern.stan0206 9 7 23
	it "Doesn't trigger on plain newtype" $
	noObservation AntiPattern.stan0206 12
	it "Doesn't trigger on a record newtype" $
	noObservation AntiPattern.stan0206 14
	it "Doesn't trigger on strict sum type field among many fields" $
	noObservation AntiPattern.stan0206 19
	it "Finds lazy field in a sum type constructor with multiple fields" $
	checkObservation AntiPattern.stan0206 20 9 12
	it "Doesn't trigger on a single strict sum type field" $
	noObservation AntiPattern.stan0206 21
	it "Finds single lazy field in a sum type with multiple constructors" $
	checkObservation AntiPattern.stan0206 22 11 15
	it "Doesn't trigger on forall wo constraint with 1 var" $
	noObservation AntiPattern.stan0206 25
	it "Finds single lazy field after forall wo constraint with 1 var" $
	checkObservation AntiPattern.stan0206 26 16 17
	it "Doesn't trigger on forall wo constraint with 2 var" $
	noObservation AntiPattern.stan0206 27
	it "Finds single lazy field after forall wo constraint with 2 var" $
	checkObservation AntiPattern.stan0206 28 17 18
	it "Doesn't trigger on strict field after forall wo constraint with 2 var" $
	noObservation AntiPattern.stan0206 29
	it "Doesn't trigger on forall constraint with 1 var" $
	noObservation AntiPattern.stan0206 30
	it "Finds single lazy field after forall constraint with 1 var" $
	checkObservation AntiPattern.stan0206 31 17 18
	it "Doesn't trigger on forall constraint with 2 var" $
	noObservation AntiPattern.stan0206 32
	it "Finds single lazy field after forall constraint with 2 var" $
	checkObservation AntiPattern.stan0206 33 17 18
	it "Doesn't trigger on strict field after forall constraint with 2 var" $
	noObservation AntiPattern.stan0206 34
	it "Doesn't trigger on forall constraint on record type" $
	noObservation AntiPattern.stan0206 36
	it "Findsa lazy Int field after forall constraint on record type" $
	checkObservation AntiPattern.stan0206 38 7 19
	it "Finds a lazy constrainted field after forall on record type" $
	checkObservation AntiPattern.stan0206 39 7 17
	describe "With the 'StrictData' extension" $ do
	let noObservation = noObservationAssert ["AntiPattern", "Stan0206Extensions"] analysis
	it "Doesn't trigger on a simple record field" $
	noObservation AntiPattern.stan0206 9
	it "Doesn't trigger on explicitly lazy field" $
	noObservation AntiPattern.stan0206 10
	it "Doesn't trigger on plain data type" $
	noObservation AntiPattern.stan0206 13

[codygman]

I just tried reproducing from NixOS and couldn't.... might be something machine specific for @goolord. We'll have to take a look.

Reproduction attempt

[cody@nixos:/tmp/stan-stricdata-minimal]$ PAGER=cat git log --oneline
be50c35 (HEAD -> master, origin/master, origin/HEAD) first commit

[cody@nixos:/tmp/stan-stricdata-minimal]$ nix-shell -p pkgs.haskell.compiler.ghc883 --run "cabal build"
Up to date
[cody@nixos:/tmp/stan-stricdata-minimal]$ ~/code/stan/.stack-work/install/x86_64-linux-nix/21146bec5e0de701f60e46f363fd96007d3477ee1dd9fe80145db1c049dace60/8.8.3/bin/stan report
  ⓘ Checking environment variables and CLI arguments for default configurations file usage...
    Fiasco:
      * [Error  ] No STAN_USE_DEFAULT_CONFIG Env Variable is set
      * [Error  ] No CLI option specified for no-default

  ⓘ The following Configurations are used:

    Result:
   
    With the following warnings:
      * [Warning] TOML Configurations file doesn't exist: /tmp/stan-stricdata-minimal/.stan.toml
      * [Warning] TOML Configurations file doesn't exist: /home/cody/.stan.toml
      * [Warning] No CLI option specified for: checks
      * [Warning] configChecks is set through the source: Default
      * [Warning] TOML Configurations file doesn't exist: /tmp/stan-stricdata-minimal/.stan.toml
      * [Warning] TOML Configurations file doesn't exist: /home/cody/.stan.toml
      * [Warning] No CLI option specified for: remove
      * [Warning] configRemoved is set through the source: Default
      * [Warning] TOML Configurations file doesn't exist: /tmp/stan-stricdata-minimal/.stan.toml
      * [Warning] TOML Configurations file doesn't exist: /home/cody/.stan.toml
      * [Warning] No CLI option specified for: ignore
      * [Warning] configIgnored is set through the source: Default

  ⓘ Using the following .cabal file: /tmp/stan-stricdata-minimal/stan-stricdata-minimal.cabal

  ✔ All clean! Stan did not find any observations at the moment.

           Stan's Summary:
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┓
┃ Analysed modules             ┃ 1      ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Analysed Lines of Code       ┃ 11     ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Total Haskell2010 extensions ┃ 1      ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Total SafeHaskell extensions ┃ 0      ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Total checked inspections    ┃ 44     ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Total found observations     ┃ 0      ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Total ignored observations   ┃ 0      ┃
┣━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╋━━━━━━━━┫
┃ Project health               ┃ 100%   ┃
┗━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┻━━━━━━━━┛

  ⓘ Report is generated here -> stan.html

[goolord]

I deleted the .hie directory and I can't reproduce this anymore. I don't know what was causing this, it persisted through rebuilds yesterday, but I think it's safe to just close this. Thanks for looking into it 👍

[vrom911]

No worries! Let us know if you see some issues again 🙂

[srid]

I can reliably reproduce this. Here's a repro using Nix:

srid/haskell-template#147

EDIT: In this case, fix was srid/haskell-flake#386