Detect and use CM files for smlnj (#884)

* Detect and use CM files for smlnj

* Split into two checkers

- one for CM projects
- one for single SML files

* Fix some typos

* Fix error caught by writing tests

We want to actually use `glob` to search in paths upwards from us.
(Previously we were just searching in the current directory every time!)

* Fix errors from former test run

* Write tests for GetCmFile and GetExecutableSmlnj

* Typo in 'smlnj/' fixture filenames

11 files changed, 205 insertions, 49 deletions

diff --git a/ale_linters/sml/smlnj.vim b/ale_linters/sml/smlnj.vim
index 4acfc9e6..f15579ea 100644
--- a/ale_linters/sml/smlnj.vim
+++ b/ale_linters/sml/smlnj.vim
@@ -1,47 +1,9 @@
-" Author: Paulo Alem    <paulo.alem@gmail.com>
-" Description: Rudimentary SML checking with smlnj compiler
-
-function! ale_linters#sml#smlnj#Handle(buffer, lines) abort
-    " Try to match basic sml errors
-
-    let l:out = []
-    let l:pattern = '^.*\:\([0-9\.]\+\)\ \(\w\+\)\:\ \(.*\)'
-    let l:pattern2 = '^.*\:\([0-9]\+\)\.\?\([0-9]\+\).* \(\(Warning\|Error\): .*\)'
-
-    for l:line in a:lines
-        let l:match2 = matchlist(l:line, l:pattern2)
-
-        if len(l:match2) != 0
-          call add(l:out, {
-          \   'bufnr': a:buffer,
-          \   'lnum': l:match2[1] + 0,
-          \   'col' : l:match2[2] - 1,
-          \   'text': l:match2[3],
-          \   'type': l:match2[3] =~# '^Warning' ? 'W' : 'E',
-          \})
-          continue
-        endif
-
-        let l:match = matchlist(l:line, l:pattern)
-
-        if len(l:match) != 0
-          call add(l:out, {
-          \   'bufnr': a:buffer,
-          \   'lnum': l:match[1] + 0,
-          \   'text': l:match[2] . ': ' . l:match[3],
-          \   'type': l:match[2] is# 'error' ? 'E' : 'W',
-          \})
-          continue
-        endif
-
-    endfor
-
-    return l:out
-endfunction
+" Author: Paulo Alem <paulo.alem@gmail.com>, Jake Zimmerman <jake@zimmerman.io>
+" Description: Single-file SML checking with SML/NJ compiler
 
 call ale#linter#Define('sml', {
 \   'name': 'smlnj',
-\   'executable': 'sml',
+\   'executable_callback': 'ale#handlers#sml#GetExecutableSmlnjFile',
 \   'command': 'sml',
-\   'callback': 'ale_linters#sml#smlnj#Handle',
+\   'callback': 'ale#handlers#sml#Handle',
 \})
diff --git a/ale_linters/sml/smlnj_cm.vim b/ale_linters/sml/smlnj_cm.vim
new file mode 100644
index 00000000..93cee63c
--- /dev/null
+++ b/ale_linters/sml/smlnj_cm.vim
@@ -0,0 +1,25 @@
+" Author: Jake Zimmerman <jake@zimmerman.io>
+" Description: SML checking with SML/NJ Compilation Manager
+
+" Let user manually set the CM file (in case our search for a CM file is
+" ambiguous and picks the wrong one)
+"
+" See :help ale-sml-smlnj for more information.
+call ale#Set('sml_smlnj_cm_file', '*.cm')
+
+function! ale_linters#sml#smlnj_cm#GetCommand(buffer) abort
+    let l:cmfile = ale#handlers#sml#GetCmFile(a:buffer)
+    return 'sml -m ' . l:cmfile . ' < /dev/null'
+endfunction
+
+" Using CM requires that we set "lint_file: 1", since it reads the files
+" from the disk itself.
+call ale#linter#Define('sml', {
+\   'name': 'smlnj-cm',
+\   'executable_callback': 'ale#handlers#sml#GetExecutableSmlnjCm',
+\   'lint_file': 1,
+\   'command_callback': 'ale_linters#sml#smlnj_cm#GetCommand',
+\   'callback': 'ale#handlers#sml#Handle',
+\})
+
+" vim:ts=4:sts=4:sw=4
diff --git a/autoload/ale/handlers/sml.vim b/autoload/ale/handlers/sml.vim
new file mode 100644
index 00000000..822a2eff
--- /dev/null
+++ b/autoload/ale/handlers/sml.vim
@@ -0,0 +1,87 @@
+" Author: Jake Zimmerman <jake@zimmerman.io>
+" Description: Shared functions for SML linters
+
+function! ale#handlers#sml#GetCmFile(buffer) abort
+    let l:pattern = ale#Var(a:buffer, 'sml_smlnj_cm_file')
+    let l:as_list = 1
+
+    let l:cmfile = ''
+    for l:path in ale#path#Upwards(expand('#' . a:buffer . ':p:h'))
+        let l:results = glob(l:path . '/' . l:pattern, 0, l:as_list)
+        if len(l:results) > 0
+            " If there is more than one CM file, we take the first one
+            " See :help ale-sml-smlnj for how to configure this.
+            let l:cmfile = l:results[0]
+        endif
+    endfor
+
+    return l:cmfile
+endfunction
+
+" Only one of smlnj or smlnj-cm can be enabled at a time.
+" executable_callback is called before *every* lint attempt
+function! s:GetExecutable(buffer, source) abort
+    if ale#handlers#sml#GetCmFile(a:buffer) is# ''
+        " No CM file found; only allow single-file mode to be enabled
+        if a:source is# 'smlnj-file'
+            return 'sml'
+        elseif a:source is# 'smlnj-cm'
+            return ''
+        endif
+    else
+        " Found a CM file; only allow cm-file mode to be enabled
+        if a:source is# 'smlnj-file'
+            return ''
+        elseif a:source is# 'smlnj-cm'
+            return 'sml'
+        endif
+    endif
+endfunction
+
+function! ale#handlers#sml#GetExecutableSmlnjCm(buffer) abort
+    return s:GetExecutable(a:buffer, 'smlnj-cm')
+endfunction
+function! ale#handlers#sml#GetExecutableSmlnjFile(buffer) abort
+    return s:GetExecutable(a:buffer, 'smlnj-file')
+endfunction
+
+function! ale#handlers#sml#Handle(buffer, lines) abort
+    " Try to match basic sml errors
+    " TODO(jez) We can get better errorfmt strings from Syntastic
+
+    let l:out = []
+    let l:pattern = '^.*\:\([0-9\.]\+\)\ \(\w\+\)\:\ \(.*\)'
+    let l:pattern2 = '^.*\:\([0-9]\+\)\.\?\([0-9]\+\).* \(\(Warning\|Error\): .*\)'
+
+    for l:line in a:lines
+        let l:match2 = matchlist(l:line, l:pattern2)
+
+        if len(l:match2) != 0
+          call add(l:out, {
+          \   'bufnr': a:buffer,
+          \   'lnum': l:match2[1] + 0,
+          \   'col' : l:match2[2] - 1,
+          \   'text': l:match2[3],
+          \   'type': l:match2[3] =~# '^Warning' ? 'W' : 'E',
+          \})
+          continue
+        endif
+
+        let l:match = matchlist(l:line, l:pattern)
+
+        if len(l:match) != 0
+          call add(l:out, {
+          \   'bufnr': a:buffer,
+          \   'lnum': l:match[1] + 0,
+          \   'text': l:match[2] . ': ' . l:match[3],
+          \   'type': l:match[2] is# 'error' ? 'E' : 'W',
+          \})
+          continue
+        endif
+
+    endfor
+
+    return l:out
+endfunction
+
+" vim:ts=4:sts=4:sw=4
diff --git a/doc/ale-sml.txt b/doc/ale-sml.txt
new file mode 100644
index 00000000..cc8d6794
--- /dev/null
+++ b/doc/ale-sml.txt
@@ -0,0 +1,36 @@
+===============================================================================
+ALE SML Integration                                           *ale-sml-options*
+
+===============================================================================
+smlnj                                                           *ale-sml-smlnj*
+                                                             *ale-sml-smlnj-cm*
+
+There are two SML/NJ powered checkers:
+
+- one using Compilation Manager that works on whole projects, but requires you
+  to save before errors show up
+- one using the SML/NJ REPL that works as you change the text, but might fail
+  if your project can only be built with CM.
+
+We dynamically select which one to use based whether we find a `*.cm` file at
+or above the directory of the file being checked. Only one checker (`smlnj`,
+`smlnj-cm`) will be enabled at a time.
+
+-------------------------------------------------------------------------------
+
+g:ale_sml_smlnj_cm_file                               *g:ale_sml_smlnj_cm_file*
+                                                      *b:ale_sml_smlnj_cm_file*
+  Type: |String|
+  Default: `'*.cm'`
+
+  By default, ALE will look for a `*.cm` file in your current directory,
+  searching upwards. It stops when it finds at least one `*.cm` file (taking
+  the first file if there are more than one).
+
+  Change this option (in the buffer or global scope) to control how ALE finds
+  CM files. For example, to always search for a CM file named `sandbox.cm`:
+>
+      let g:ale_sml_smlnj_cm_file = 'sandbox.cm'
+
+===============================================================================
+  vim:tw=78:ts=2:sts=2:sw=2:ft=help:norl:
diff --git a/doc/ale.txt b/doc/ale.txt
index fe91d4da..a8b30213 100644
--- a/doc/ale.txt
+++ b/doc/ale.txt
@@ -119,6 +119,8 @@ CONTENTS                                                         *ale-contents*
     sh....................................|ale-sh-options|
       shell...............................|ale-sh-shell|
       shellcheck..........................|ale-sh-shellcheck|
+    sml...................................|ale-sml-options|
+      smlnj...............................|ale-sml-smlnj|
     spec..................................|ale-spec-options|
       rpmlint.............................|ale-spec-rpmlint|
     stylus................................|ale-stylus-options|
diff --git a/test/handler/test_sml_handler.vader b/test/handler/test_sml_handler.vader
index 26c85710..f711cc9c 100644
--- a/test/handler/test_sml_handler.vader
+++ b/test/handler/test_sml_handler.vader
@@ -1,6 +1,3 @@
-Before:
-  runtime ale_linters/sml/smlnj.vim
-
 Execute (Testing on EOF error):
   AssertEqual [
   \ {
@@ -11,7 +8,7 @@ Execute (Testing on EOF error):
   \   'text': 'Error: syntax error found at EOF',
   \ },
   \],
-  \ ale_linters#sml#smlnj#Handle(42, [
+  \ ale#handlers#sml#Handle(42, [
   \ "Standard ML of New Jersey v110.78 [built: Thu Jul 23 11:21:58 2015]",
   \ "[opening a.sml]",
   \ "a.sml:2.16 Error: syntax error found at EOF",
@@ -35,7 +32,7 @@ Execute (Testing if the handler can handle multiple errors on the same line):
   \   'text': 'Error: unbound variable or constructor: wow',
   \ },
   \],
-  \ ale_linters#sml#smlnj#Handle(42, [
+  \ ale#handlers#sml#Handle(42, [
   \ "Standard ML of New Jersey v110.78 [built: Thu Jul 23 11:21:58 2015]",
   \ "[opening test.sml]",
   \ "a.sml:1.6-1.10 Error: can't find function arguments in clause",
@@ -61,7 +58,7 @@ Execute (Testing rarer errors):
   \   'text': "Error: value type in structure doesn't match signature spec",
   \ },
   \],
-  \ ale_linters#sml#smlnj#Handle(42, [
+  \ ale#handlers#sml#Handle(42, [
   \ "Standard ML of New Jersey v110.78 [built: Thu Jul 23 11:21:58 2015]",
   \ "[opening test.sml]",
   \ "a.sml:5.19 Error: syntax error found at ID",
@@ -80,7 +77,7 @@ Execute (Testing a warning):
   \   'text': "Warning: match nonexhaustive",
   \ },
   \],
-  \ ale_linters#sml#smlnj#Handle(42, [
+  \ ale#handlers#sml#Handle(42, [
   \ "Standard ML of New Jersey v110.78 [built: Thu Jul 23 11:21:58 2015]",
   \ "[opening a.sml]",
   \ "a.sml:4.5-4.12 Warning: match nonexhaustive",
diff --git a/test/smlnj/cm/foo.sml b/test/smlnj/cm/foo.sml
new file mode 100644
index 00000000..e69de29b
--- /dev/null
+++ b/test/smlnj/cm/foo.sml
diff --git a/test/smlnj/cm/path/to/bar.sml b/test/smlnj/cm/path/to/bar.sml
new file mode 100644
index 00000000..e69de29b
--- /dev/null
+++ b/test/smlnj/cm/path/to/bar.sml
diff --git a/test/smlnj/cm/sources.cm b/test/smlnj/cm/sources.cm
new file mode 100644
index 00000000..e69de29b
--- /dev/null
+++ b/test/smlnj/cm/sources.cm
diff --git a/test/smlnj/file/qux.sml b/test/smlnj/file/qux.sml
new file mode 100644
index 00000000..e69de29b
--- /dev/null
+++ b/test/smlnj/file/qux.sml
diff --git a/test/test_sml_command.vader b/test/test_sml_command.vader
new file mode 100644
index 00000000..5ce8a313
--- /dev/null
+++ b/test/test_sml_command.vader
@@ -0,0 +1,47 @@
+Before:
+  runtime ale_linters/sml/sml.vim
+  runtime ale_linters/sml/smlnj.vim
+  call ale#test#SetDirectory('/testplugin/test')
+
+After:
+  call ale#test#RestoreDirectory()
+  call ale#linter#Reset()
+
+# ----- GetCmFile -----
+
+Execute(smlnj finds CM file if it exists):
+  call ale#test#SetFilename('smlnj/cm/foo.sml')
+
+  AssertEqual '/testplugin/test/smlnj/cm/sources.cm', ale#handlers#sml#GetCmFile(bufnr('%'))
+
+Execute(smlnj finds CM file by searching upwards):
+  call ale#test#SetFilename('smlnj/cm/path/to/bar.sml')
+
+  AssertEqual '/testplugin/test/smlnj/cm/sources.cm', ale#handlers#sml#GetCmFile(bufnr('%'))
+
+Execute(smlnj returns '' when no CM file found):
+  call ale#test#SetFilename('smlnj/file/qux.sml')
+
+  AssertEqual '', ale#handlers#sml#GetCmFile(bufnr('%'))
+
+# ----- GetExecutableSmlnjCm & GetExecutableSmlnjFile -----
+
+Execute(CM-project mode enabled when CM file found):
+  call ale#test#SetFilename('smlnj/cm/foo.sml')
+
+  AssertEqual 'sml', ale#handlers#sml#GetExecutableSmlnjCm(bufnr('%'))
+
+Execute(single-file mode disabled when CM file found):
+  call ale#test#SetFilename('smlnj/cm/foo.sml')
+
+  AssertEqual '', ale#handlers#sml#GetExecutableSmlnjFile(bufnr('%'))
+
+Execute(CM-project mode disabled when CM file not found):
+  call ale#test#SetFilename('smlnj/file/qux.sml')
+
+  AssertEqual '', ale#handlers#sml#GetExecutableSmlnjCm(bufnr('%'))
+
+Execute(single-file mode enabled when CM file found):
+  call ale#test#SetFilename('smlnj/file/qux.sml')
+
+  AssertEqual 'sml', ale#handlers#sml#GetExecutableSmlnjFile(bufnr('%'))