#198 Deal with Octave script-global functions

* In Octave mode this is now supported

* In MATLAB mode this now produces a very specific error

Closes #198

CHANGELOG.md

@@ -46,6 +46,12 @@ Not quite compatible with Octave yet. See #43 [octave support](https://github.co
   be unexpectedly shared between different directory trees, this is
   now fixed and is working as documented.
 
+* Support Octave script-global functions in `--octave` mode. These are
+  function definitions that are sprinkled in the middle of a script
+  file, instead of appearing at the end. Also vastly improve the error
+  message that appears when encountering this construct in MATLAB
+  mode.
+
 ### 0.9.15
 
 * Added [documentation](https://florianschanda.github.io/miss_hit/configuration.html#pre-commit) on how to use MISS_HIT through pre-commit hooks.

miss_hit_core/m_lexer.py

@@ -82,6 +82,12 @@ def __init__(self, filename, blockname=None):
                os.path.dirname(os.path.abspath(filename))).startswith("@")
         # Make a note if this file resides inside a @ directory
 
+        self.octave_mode = False
+        # If set to true, also deal with Octave's extensions to
+        # MATLAB.
+        #
+        # Note that this is highly incomplete right now.
+
     @abstractmethod
     def token(self):
         pass
@@ -177,12 +183,6 @@ def __init__(self, mh, content, filename, blockname=None):
         # pragmas. Hence there is an option to turn this off. It is
         # also turned off in config file mode.
 
-        self.octave_mode = False
-        # If set to true, also deal with Octave's extensions to
-        # MATLAB.
-        #
-        # Note that this is highly incomplete right now.
-
         self.block_comment = 0
         # We're in a block comment and are ignore almost everything
         # (except a closing block comment). This is a number because
@@ -1219,6 +1219,7 @@ def __init__(self, lexer, cfg):
         self.mh = lexer.mh
         self.lines = lexer.line_count()
         self.comment_char = lexer.comment_char
+        self.octave_mode = lexer.octave_mode
 
         while True:
             tok = lexer.token()

miss_hit_core/m_parser.py

@@ -629,8 +629,18 @@ def parse_file(self):
                                   l_pragmas + l_more_pragmas)
         elif self.peek("KEYWORD", "classdef") or self.lexer.in_class_directory:
             cunit = self.parse_class_file(l_pragmas)
+        elif self.lexer.octave_mode:
+            cunit = self.parse_octave_script_file(l_pragmas)
         else:
-            cunit = self.parse_script_file(l_pragmas)
+            cunit = self.parse_matlab_script_file(l_pragmas)
+
+            # Special detection of Octave inline globals in MATLAB
+            # mode (to improve the messages produced in #198).
+            if not self.peek_eof() and cunit.l_functions:
+                self.mh.error(cunit.l_functions[0].loc(),
+                              "script-global functions are an Octave-specific"
+                              " feature; move your functions to the end of"
+                              " the script file or use the --octave mode")
 
         if self.debug_tree:
             cunit.debug_parse_tree()
@@ -639,9 +649,14 @@ def parse_file(self):
 
         return cunit
 
-    def parse_script_file(self, l_pragmas):
+    def parse_matlab_script_file(self, l_pragmas):
+        # A MATLAB script file is a bunch of statements, followed by
+        # zero or more private functions. This is distinct from Octave
+        # script files, which allow you to sprinkle functions anywhere
+        # in the top-level context.
+
         # At least in MATLAB 2017b script files cannot use the
-        # non-ended chained functions
+        # non-ended chained functions.
 
         self.functions_require_end = True
         statements = []
@@ -662,6 +677,42 @@ def parse_script_file(self, l_pragmas):
                            l_functions,
                            l_pragmas + l_more_pragmas)
 
+    def parse_octave_script_file(self, l_pragmas):
+        # An Octave script file is very similar to a MATLAB one,
+        # except they allow you to put functions everywhere (as long
+        # as it's not the first thing, because then you have a
+        # function file again). The semantics are weird as well,
+        # because all functions encountered this way are made global,
+        # as if they were in their own function files.
+
+        self.functions_require_end = True
+        statements = []
+        l_functions = []
+        l_more_pragmas = []
+
+        if not self.peek_eof():
+            # Script files _have_ to start with a statment
+            statements.append(self.parse_statement())
+
+        # Otherwise, in Octave, script files are a sequence of
+        # statements intermixed with (global) function definitions
+        while not self.peek_eof():
+            if self.peek("KEYWORD", "function"):
+                l_functions.append(self.parse_function_def())
+            elif self.peek("KEYWORD", "pragma"):
+                l_more_pragmas.append(self.parse_annotation_pragma())
+            else:
+                statements.append(self.parse_statement())
+
+        return Script_File(os.path.basename(self.lexer.filename),
+                           os.path.dirname(
+                               os.path.abspath(self.lexer.filename)),
+                           self.lexer.get_file_loc(),
+                           self.lexer.line_count(),
+                           Sequence_Of_Statements(statements),
+                           l_functions,
+                           l_pragmas + l_more_pragmas)
+
     def parse_class_file(self, l_pragmas):
         self.functions_require_end = True
 

tests/lint/octave_global_functions/README.md

@@ -0,0 +1,3 @@
+Octave has very confusing semantics when it comes to functions in
+*script* files. These are made globally available, as if they were in
+their own `m` file; but only if the script is run first.

tests/lint/octave_global_functions/expected_out.txt

@@ -0,0 +1,2 @@
+=== PLAIN MODE ===
+MISS_HIT Lint Summary: 7 file(s) analysed, everything seems fine

tests/lint/octave_global_functions/foo.m

@@ -0,0 +1,7 @@
+1;
+
+function x = potato(y)
+    x = y + 1;
+end
+
+disp('Hello world!');

tests/lint/octave_global_functions/miss_hit.cfg

@@ -0,0 +1 @@
+octave: true

tests/lint/octave_global_functions/test_1.m

@@ -0,0 +1,6 @@
+foo;
+disp(potato(12));
+
+% Not an error, because foo is a bizarre octave extension that makes
+% functions in the middle go global. Specifically this happens only
+% for _script_ files

tests/lint/octave_global_functions/test_1_b.m

@@ -0,0 +1,3 @@
+disp(potato(12));
+
+% Error, because potato (from the script file foo) is not visible.

tests/lint/octave_global_functions/test_2.m

@@ -0,0 +1,3 @@
+disp(wibble(5));
+
+% Should print 11.

tests/lint/octave_global_functions/test_3.m

@@ -0,0 +1,3 @@
+disp(wobble(5));
+
+% Error, because wobble is not visible/private

tests/lint/octave_global_functions/test_4.m

@@ -0,0 +1,5 @@
+disp(wibble(5));
+% Prints 11
+
+disp(wobble(5));
+% Error, because wobble is still not visible/private

tests/lint/octave_global_functions/wibble.m

@@ -0,0 +1,7 @@
+function rv = wibble(x)
+    rv = wobble(x) + 1;
+end
+
+function rv = wobble(x)
+    rv = x * 2;
+end

tests/parser/scripts/bar.m.out

@@ -1,4 +1,4 @@
-In bar.m, line 7
-| kittens = 42; % Illegal
-| ^^^^^^^ error: expected end of file, found IDENTIFIER instead
+In bar.m, line 3
+| function y = foo(x)
+|              ^^^ error: script-global functions are an Octave-specific feature; move your functions to the end of the script file or use the --octave mode
 MISS_HIT Debug Summary: 1 file(s) analysed, 1 error(s)

tests/style/bug_199_parse_errors_and_style/expected_out.html

@@ -13,7 +13,7 @@ <h1>Issues identified</h1>
 <section>
 <h2>test_1.m</h2>
 <div class="message"><a href="matlab:opentoline('test_1.m', 1, 45)">test_1.m: line 1:</a> style: No copyright notice found in header</div>
-<div class="message"><a href="matlab:opentoline('test_1.m', 16)">test_1.m: line 16:</a> error: expected end of file, found IDENTIFIER instead</div>
+<div class="message"><a href="matlab:opentoline('test_1.m', 5, 14)">test_1.m: line 5:</a> error: script-global functions are an Octave-specific feature; move your functions to the end of the script file or use the --octave mode</div>
 <h2>test_2.m</h2>
 <div class="message"><a href="matlab:opentoline('test_2.m', 1, 45)">test_2.m: line 1:</a> style: No copyright notice found in header</div>
 <div class="message"><a href="matlab:opentoline('test_2.m', 16, 7)">test_2.m: line 16:</a> style: end statement with a semicolon</div>

tests/style/bug_199_parse_errors_and_style/expected_out.txt

@@ -3,9 +3,9 @@ test_1.m: error: file is not auto-fixed because it contains parse errors
 In test_1.m, line 1
 | % Taken from issue #199 (by alvinseville7cf)
 |                                             ^ style: No copyright notice found in header
-In test_1.m, line 16
-| n = 4;
-| ^ error: expected end of file, found IDENTIFIER instead
+In test_1.m, line 5
+| function z = f(x, a)
+|              ^ error: script-global functions are an Octave-specific feature; move your functions to the end of the script file or use the --octave mode
 In test_2.m, line 1
 | % Taken from issue #199 (by alvinseville7cf)
 |                                             ^ style: No copyright notice found in header