Trouble making a functionList parser for MATLAB
-
This should take care of displaying the classes as branches in the tree, haven’t figured out the block comment problem yet i.e.
function_no_return_C2( val )incorrectly being displayed.<parser displayName="Matlab Node" id ="matlab" commentExpr="(?x) # Utilize inline comments (see `RegEx - Pattern Modifiers`) (?s:%\{.*%\}) # Multi Line Comment | (?m-s:%.*$) # Single Line Comment " > <classRange mainExpr="(?x) # Utilize inline comments (see `RegEx - Pattern Modifiers`) (?m) # ^ and $ match at line breaks ^classdef\h+ # start-of-class indicator at start-of-line (?s:.*?) # whatever, until... ^end # end-of-class indicator at start-of-line " > <className> <nameExpr expr="(?x) # Utilize inline comments (see `RegEx - Pattern Modifiers`) classdef\h+ # start-of-class indicator \K # keep the text matched so far, out of the overall match [A-Za-z]\w* # valid identifier i.e. class name " /> </className> <function mainExpr="(?x) # Utilize inline comments (see `RegEx - Pattern Modifiers`) (?m) # ^ and $ match at line breaks ^(?'INDENT'\h{2,}) # at least two whitespaces at start-of-line; store for back-reference function\h+ # start-of-function indicator \K # keep the text matched so far, out of the overall match (?s:.*?) # whatever, until... ^\k'INDENT'end\b # end-of-function indicator with equal indent " > <functionName> <funcNameExpr expr="(?x) # Utilize inline comments (see `RegEx - Pattern Modifiers`) (?m-s) # ^ and $ match at line breaks; dot does not .* # whatever, up till... (?=;|$) # semi-colon or end-of-line " /> </functionName> </function> </classRange> <function mainExpr="(?x) # Utilize inline comments (see `RegEx - Pattern Modifiers`) (?m-s) # ^ and $ match at line breaks; dot does not ^\h* # optional leading whitespace at start-of-line function\h+ # start-of-function indicator \K # keep the text matched so far, out of the overall match .*$ # match the rest of the line " /> </parser>@Claudia-Frank created a Python script to be used with the PythonScript plugin. It enables you to test a regular expression in Notepad++ jumping through every match.
I’ve used RegEx Tester until I bought a license for RegexBuddy.For “class” definitions the main expression has to match the complete class i.e. including body.
The parser first determines all the comment regions. Skipping these comment regions it searches for regions matching the class expression. Any remaining regions (i.e. not matching comment or class) are searched for (global) functions (or methods defined outside a class).
Each class region is searched for its class name and methods. -
just in case you are looking for the script.
Here is the more user-friendly copy and paste version (not splitted)Cheers
Claudia -
Thank you very much for your prompt and detailed answer. Your code works perfectly. Together with your explanation, it helped me better understand the working of functionList, which in turn effectively helped me to spot my mistakes.
-
Thank you for the link.
How can I earn two points to be allowed to send two post directly after each other.
Cheers
Per -
@MAPJe71
Will you take care of creating a functionlist PR for matlab out of this for N++? -
@Per-Isakson glad to hear it works!
Your code works perfectly.
Does this mean the “block comment problem” is no longer an issue for you?
@chcg Yes, I intend to include it in PR “Function List Update 3”.
-
The “block comment problem” is still there, but it is NOT an issue to me. This problem exists even in the MATLAB IDE. Commenting out code is an anti-pattern. Thus, put it near the bottom of your priority list.
“Yes, I intend to include it in PR “Function List Update 3””
I like that! However, this MATLAB <parser> is an exercise. I guess, it will not be very helpful in serious work. Let me make a more complete version.
A “full” <parser> as the first step was obviously a too big an undertaking for me. Now I intend to make a “full” <parser>. However, I see some problems.
- MATLAB syntax allows continuation lines. Any line may be split into two or more lines. I have failed to handle this with MATLAB regular expressions. I’m not tempted to try again. (Minor issue.)
- MATLAB syntax allows definition of class methods (/functions) in separate files. I don’t use that, but it is frequently used by The Mathworks and many users. I cannot see any possibility to handle this.
- Indentation is not significant;
endis used to close nearly all blocks; and ordinary functions may be defined at the end of class definition files. Together these three pose a major challenge. (My^\h{2,}was a quick and dirty way to handle this.)
I think, I will use the functionList in two different scenarios, which ask for different versions of the <parser>:
- Navigation in code, which I know fairly well, e.g. my own old code. In this case I would prefer the functionList to show names only, which is cleaner and fits in a narrow plane.
- Analysis of code, which I download from the Internet. In this case I think the full signature is more useful. (Like in my current exercise.)
Conclusion: I’ll be back with a new version of the <parser>.
-
The “block comment problem” … put it near the bottom of your priority list.
For me it’s not a MATLAB parser problem only it’s more a FunctionList engine problem …
- why does it work for other parsers (or at least seems to)?
- why does the comment expression work correctly with MATLAB source in RegexBuddy and with ClaudiaFrank’s RegexTester script in Notepad++ but apparently not with Notepad++ Function List engine?
- why does the comment expression work correctly when the
classRangepart of the parser is omitted?
this MATLAB <parser> is an exercise … Let me make a more complete version.
I expect “Function List Update 3” PR not to be the last one, so it’s possible to include it in a later update/PR or you could do your own PR.
Either way I’d appreciate it when you keep me in the loop :)MATLAB syntax allows continuation lines.
Could you give an example?
MATLAB syntax allows definition of class methods (/functions) in separate files.
Not supported as FunctionList operates per file.
Indentation is not significant …
That will make it a challenge to determine the
endof a class definition! Determining theendof a function/method definition may not be necessary.I’ll be back with a new version of the <parser>.
Looking forward to it!
-
The “block comment problem” disappeared as a result of changes of the <parser>, which have no obvious coupling to comments. Strange!
MATLAB syntax allows continuation lines. Could you give an example?
function ... WeirdUseOfContinuationLine ( ... str ) disp( str ) % This runs as expexted. endI guess this could be handled with regex, but I will not try.
-
The “block comment problem” disappeared as a result of changes of the <parser>
Strange indeed.
I presume you applied the changes, could you post the updated parser (even if it’s not the final one)?I guess this could be handled with regex
Possibly.
but I will not try.
Wise, keep it as challenge for an other revision :-)
-
Thank you for your answers, especially on Nested Pairs and recursion. Now I have a <parser>, which doesn’t rely on the indentation of the Matlab code and produces correct results for my first set of test files. However, keywords embedded in comments and in string values make the <parser> fail.
A new <parser> and some test files
<parser id = "matlab" displayName = "Matlab Node" commentExpr = "(?x) # Utilize inline comments (see `RegEx - Pattern Modifiers`) (?s:%\{.*%\}) # Multi Line Comment |(?m-s:\h*%.*$) # Single Line Comment " > <classRange mainExpr = "(?x)(?s) # dot matches new line (?-i) # case sensitive ( # --- open 1st group \b # word boundary ( # --- open 2nd group classdef # keywords that open a | e(?:numeration|vents) # Balanced Construct | f(?:or|unction) # that is closed by | if # 'end'. | methods # | p(?:arfor|roperties) # | switch # | try # | while # ) # --- close 2nd group \b # word boundary ) # --- close 1st group ( # open 3rd capturing group (?: # open non-capturing group (?! # negative look-ahead; if not a keyword \b # word boundary ( # --- open 4th group classdef # keywords, of the 2nd group + 'end', | e(?:numeration|vents) # which must NOT occur, | f(?:or|unction) # at ANY position, of the | if # present sub-block, till | methods # its associated 'end' closing word | p(?:arfor|roperties) # | switch # | try # | while # | end # ) # --- close 4th group \b # word boundary ). # then one character )+ # repeat | # until keyword found (?R) # recurse RE from start )+ # repeat 3rd group \bend\b # " > <className> <nameExpr expr = "(?x) # ok, 10:07 2017-03-25 ^\h* # optional whitespace classdef # start-of-class indicator \h+ # whitespace ( # \([^\)]*?\) # (ClassAttributes) \h+ # whitespace )? # \K # discarge what's matched so far [A-Za-z]\w* # valid identifier, i.e. class name " /> </className> <function mainExpr = "(?x) # ok, 10:22 2017-03-25, [2] (?m) # ^ and $ match at line breaks ( (^\h*) # optional whitespace starting line | # or ([,;]\h*) # separator before 'function' ) function # start-of-function indicator \h+ # whitespace \K # discard what's matched so far ( # open optional left hand side, LHS ( # (\w+) # single variable name | # or (\[[\w,\h]*\]) # one or more variable names in brackets ) # \h*=\h* # )? # close optional left hand side [A-Za-z]\w*\b # function name \h* # ( # open optional argument list (\([\h,\w~]*\)) # pair of parentheses with optional arguments | # or ( # \([\h,\w~]*? # opening parenthisis and optional arguments (?=\h*\.{3}) # and closed by continuation line ) # )? # close optional argument list " > <functionName> <funcNameExpr expr = "(?x) # ok, 12:07 2017-03-25 (?m) # ^ and $ match at line breaks (?-s) # single-line ( # open optional left hand side, LHS ((\w+)| # single variable name (\[[\w,\h]*\])) # one or more variable names in brackets \h*=\h* # )? # close optional left hand side [A-Za-z]\w*\b # function name \h* # ( # open optional argument list (\([\h,\w~]*\)) # pair of parenthisis with optional arguments | # or (\([\h,\w~]*? # lazy to avoid trailing whitespace (?=\h*\.{3})) # continuation line )? # close optional argument list " /> </functionName> </function> </classRange> <function mainExpr = "(?x) # [3] ^\h* # optional whitespace function # start-of-function indicator \h+ # whitespaces (?-s:.*) # the rest of the line " > <functionName> <nameExpr expr = "(?x) # function # start-of-function indicator \h+ # whitespace \K # discard what's matched so far ( # open optional left hand side, LHS ( # (\w+) # single variable name | # or \[[\w,\h]*\] # one or more variable names in brackets ) # \h*=\h* # )? # close optional left hand side [A-Za-z]\w* # function name \b\h* # ( # open optional input argument list \([\h,\w~]*\) # pair of parenthisis with optional arguments or | # or ( # \([\h,\w]*? # lazy to avoid trailing whitespace (?=\h*\.{3}) # ) # )? # close optional argument list " /> </functionName> </function> </parser>function_with_keyword_in_comment.m. The <parser> fails with this test file. Only the file name is displayed in Function List.function [ out, val ] = function_with_keyword_in_comment( val ) out = 17; % comment with keyword: while endclass_with_keyword_in_block_comment.m. The <parser> fails with this test file. Only the file name is displayed in Function List.classdef class_with_keyword_in_block_comment methods function no_return( this ) %{ comment with keyword: while %} a = 17; end end endclass_with_keyword_in_string.m. The <parser> fails with this test file. Only the file name is displayed in Function List.classdef class_with_keyword_in_string methods function no_return( this ) str = 'string with keyword: while'; a = 17; end end endOneLineClasses.m. The <parser> displays the expected result in Function Listclassdef OneLineClasses,methods,function OneLineMethod(~,str),disp(str),end,function [q]=OneLineMethod(str),disp(str),end,end,endUpdate the editor window first and then update the Function List
Problem: “Browsing” through open files,
Next TabandPrevious Tab, is halted for seconds by files that the <parser> fails to process.
Proposal: LetNext TabandPrevious Tabinterrupt the <parser> (if working), update the editor window, and then update Function List. While the <parser> is working the update icon could rotate to indicate ongoing work.More than one Matlab <parser>
In some use cases I would like to show the full function signature in the Function List and in other only the function name. It’s straight forward to modify the current <parser> to only display function names. However, I cannot think of a good way to switch between the two variants.
Keywords embedded in comments and in string values
The problem with keywords in comments and string values makes the current <parser> nearly worthless for practical work. I have many “for” and “end” in my comments and string values.
There is a simple solution, replace
parser.classRange.mainExprby something simple that requires that files honor the recommended format, e.g.(?m)^classdef\h+(?s:.*?)^end\b. That’s good enough for me. And I have done a useful exercise with RE. -
Update the editor window first and then update the Function List
I suspect the recursive part of
classRange.mainExpris giving the RE engine a hard time :-)More than one Matlab <parser>
Based on current Function List implementation I can think of three solutions: define a <parser> for each use-case and switch between them by …
- changing the association; this requires a restart of NP++ to take effect;
- swapping their parser id, this also requires a restart of NP++ to take effect;
- utilizing file extension association and changing the extension of the file.
Keywords embedded in comments and in string values
The trick here is to handle literate strings as comment regions.
e.g.commentExpr="(?x) # free spacing (?s:(?s:%\{.*%\}) # Multi Line Comment | (?m-s:%.*$) # Single Line Comment | (?s:\x22(?:[^\x22\x5C]|\x5C.)*\x22) # String Literal - Double Quoted | (?s:\x27(?:[^\x27\x5C]|\x5C.)*\x27) # String Literal - Single Quoted " -
@MAPJe71, thanks for your answer.
The trick here is to handle literate strings as comment regions.
Yes, but I fail to make it work. I copy&pasted your
commentExprvalue into the <parser> that I posted yesterday. None of the “keyword in comment and string” test files give the expected result in the Function List, i.e. the same result as yesterday.Speculation:
commentExprhas no effect in my <parser>.- The “block-comment-problem” is back, i.e. function_no_return_C2 is displayed
- With
classRange.mainExprusingrecursionskipping strings values and comments is needed.
-
There is definitely something weird going with the comment expression. The single line comment and both double and single quoted string literals are detected just perfectly but the multi line comment keeps failing.
Oh just noticed that the comment expression I posted has a superfluous
(?s:at the start of the “Multi Line Comment”-part. -
The single line comment and both double and single quoted string literals are detected just perfectly …
That is not how I interpret what I see om my side. I have no debugging means to see how NPP uses the value of
commentExpr. (Or I missed it.) Thus, I resort to indirect tests.With
- your new value of
commentExpr(corrected), classRange.mainExprusing recursion,- the Matlab files attached and
- Notepad++ v7.3.3 (32-bit), Build time : Mar 8 2017 - 03:37:37, Admin mode : OFF, Local Conf mode : OFF, OS : Windows 7 (64-bit), Plugins : …
Function List displays a row of dots, a document icon and the file name. The evaluation, which obviously fails, takes a few seconds.
With the keyword,
while, replaced by_whilethe Function List displays the expected results without delay.Do you see a different behavior?
I’m out of ideas, are there other test I could do?
function [ out, val ] = function_with_keyword_in_string( val ) out = 17; str = 'string with keyword: while'; endfunction [ out, val ] = function_with_keyword_in_comment( val ) out = 17; % comment with keyword: while endclassdef class_with_keyword_in_comment methods function no_return( this ) % comment with keyword: while a = 17; end end endclassdef class_with_keyword_in_string methods function no_return( this ) str = 'string with keyword: while'; a = 17; end end end - your new value of
-
This is what I get …
with the following <parser> …
<parser displayName="MATLAB - MATrix LABoratory" id ="matlab_syntax" commentExpr="(?x) # free-spacing (see `RegEx - Pattern Modifiers`) (?'MLC': # Multi Line Comment %\{ # ...start-of-comment indicator (?: # ...followed by zero or more characters [^%] # ...not start of start-of-comment indicator | %(?![%{}]) # ...not being an SLC or a start- or end-of-comment indicator )*? %\} # ...end-of-comment indicator ) | (?m-s:\x25.*$) # Single Line Comment (SLC) | (?s:\x22(?:[^\x22\x5C]|\x5C.)*\x22) # String Literal - Double Quoted | (?s:\x27(?:[^\x27\x5C]|\x5C.)*\x27) # String Literal - Single Quoted " > <classRange mainExpr ="(?x) # free-spacing (see `RegEx - Pattern Modifiers`) (?s-i) # dot matches at line breaks, case-sensitive \bclassdef\b # start-of-class indicator .*? # whatever, until... \bmethods\b # ...start-of-class-body indicator " openSymbole ="(?x) # free-spacing (see `RegEx - Pattern Modifiers`) \b # ensure leading word boundary for start-of-block indicator (?-i: # case-sensitive start-of-block indicators e(?:numeration|vents) | f(?:or|unction) | if | methods | p(?:arfor|roperties) | switch | try | while ) \b " closeSymbole="(?x) # free-spacing (see `RegEx - Pattern Modifiers`) \b # ensure leading word boundary for end-of-block indicator (?-i: # case-sensitive end # end-of-block indicator ) \b # ensure trailing word boundary for end-of-block indicator " > <className> <nameExpr expr="(?x) # free-spacing (see `RegEx - Pattern Modifiers`) \b(?-i:classdef) # case-sensitive start-of-class indicator \h+ # required whitespace (?:\([^)]*?\)\h+)? # optional class-attributes \K # discard text matched so far [A-Za-z]\w* # valid character combination for identifiers i.e. class name " /> </className> <function mainExpr="(?x) # free-spacing (see `RegEx - Pattern Modifiers`) (?ms-i) # ^, $ and dot match at line breaks, case-sensitive (?: ^ # a function can be found at start-of-line | [,;] # ...or after a separator )\h* # ...optionally followed by whitespace \K # discard text matched so far \bfunction # ensure word boundary for start-of-function indicator \s+ # required whitespace separator .*? # whatever, until... \bend\b # ...the first end-of-block indicator " > <functionName> <funcNameExpr expr="(?x) # free-spacing (see `RegEx - Pattern Modifiers`) function\s+ # start-of-function indicator (?:\.{3}\s+)? # optional continuation-line indicator \K # discard text matched so far (?: # optional return value(s) (?: \w+ # ...single variable name | \[[\h,\w]*\] # ...or one or more variable names in brackets )\h*=\h* # ...followed by a separator with optional whitespace )? [A-Za-z]\w* # valid character combination for identifiers i.e. function name \b # ensure word boundary for name (?: # optional parameter list \s* # ...optional leading whitespace \( # ...start-of-parameter-list indicator [\h,\w~]* # ...with optional parameters (?: \) # ...until end-of-parameter-list indicator | \.{3} # ...or continuation-line indicator ) )? " /> <!-- comment out the following node to display the method with its parameters --> <!-- <funcNameExpr expr="(?:(?:\w+|\[[\h,\w]*\])\h*=\h*)?[A-Za-z]\w*" /> --> </functionName> </function> </classRange> <function mainExpr="(?x) # free-spacing (see `RegEx - Pattern Modifiers`) (?ms-i) # ^, $ and dot match at line breaks, case-sensitive ^\h* # optional leading whitespace at start-of-line function\b # start-of-function indicator .*? # whatever, until... \bend\b # ...the first end-of-block indicator " > <functionName> <nameExpr expr="(?x) # free-spacing (see `RegEx - Pattern Modifiers`) function\s+ # start-of-function indicator (?:\.{3}\s+)? # optional continuation-line indicator \K # discard text matched so far (?: # optional return value(s) (?: \w+ # ...single variable name | \[[\h,\w]*\] # ...or one or more variable names in brackets )\h*=\h* # ...followed by a separator with optional whitespace )? [A-Za-z]\w* # valid character combination for identifiers i.e. function name \b # ensure word boundary for name (?: # optional parameter list \s* # ...optional leading whitespace \( # ...start-of-parameter-list indicator [\h,\w~]* # ...with optional parameters (?: \) # ...until end-of-parameter-list indicator | \.{3} # ...or continuation-line indicator ) )? " /> <!-- comment out the following node to display the method with its parameters --> <!-- <nameExpr expr="(?:(?:\w+|\[[\h,\w]*\])\h*=\h*)?[A-Za-z]\w*" /> --> </functionName> </function> </parser> -
@MAPJe71 ,
Your new parser works like a charm. It even handlesWeirdUseOfContinuationLine.m. Thank you very much! -
Sorry, I’m back with more problems. However, the result so far is good enough for me who know the limitations. Other Matlab users may think differently.
My goal is that the <parser> produces the intended result for syntactically correct Matlab files. Exceptions are acceptable for rare and smelly cases. It’s good if something is shown for syntactically incorrect Matlab files - the more the better.
The Matlab class syntax, [Class Syntax Guide], includes more than I have “communicated” with the test files in this thread.
Problem: A Matlab file allows one
classdef...endblock, which in turn may contain zero or moremethods...endblocks.classdefshall be the first keyword in the file. The current <parser> displays the functions of the firstmethods...endblock as belonging to the class, i.e. being methods, and erroneously the functions of subsequentmethods...endblocks as functions. This is illustrated by the attached m-file.I failed to modify your <parser> to show all function inside the
classdef...endblock as methods belonging to the class.classdef (Sealed) aCompleteClassWithFunction < handle properties (Access = private) Prop1 = datenum(date) end properties Prop2 end events (ListenAccess = protected) StateChanged end methods function obj = aCompleteClassWithFunction(x) obj.Prop2 = x; end end methods (Access = {?MyOtherClass}) function d = myMethod(obj) d = obj.Prop1 + x; end end end function myUtilityFcn A = 17; end -
Sigh, it was to good to be true ;-)
I’ll will look at it tomorrow. -
I’m obviously not capable to contribute effectively to the code of this <parser>:-(. I lack both regarding knowledge on regular expressions, and understanding of how the <parser>-code produces the function list.
How important is a good Matlab-<parser> to Notepad++ ? To me Notepad++ is a valuable complement to Matlab’s editor, which is sluggish and doesn’t support regular expressions, not even wildcards. I use Notepad++ daily. Notepad++ is superior when it comes to large files and browsing large code-bases. Google returns 0.18 million results for “notepad++ matlab”.
I send this post rather to help you decide than insisting on specific things being included in the <parser>. What we achieved so far is good enough for me!
Is there a mean to make the user aware of limitations in the capability of a specific <parser>. Documented limitations are more acceptable.
MLC commentExpr
The colon after ‘MLC’ is that a typo or
non-capture?Proposal: replace the MLC-code by
(?xms) (?'OMLC' (?>^\h*%\{\h*$)) .*? (?'CMLC' (?>^\h*%\}\h*$))Justification: Again Matlab’s syntax [Add Comments to Programs] is permissive
- nested multi-line-comments (candidate for documented limitation)
- single-line-comments inside MLC
- I often comments out code with both SLC and MLC, because Matlab has a feature, Find in Files, which doesn’t honor MLC (bug). MLC allows me to code-fold the comment-out code.
This should be matched by MLC
%{ some code %% Section % some other code %}This expression is affected by the block-comment-problem.
Nested functions
Matlab have nested functions. My previous <parser> showed functions nested in functions as ordinary functions and functions nested in methods broke the function list (showed only file name). Your current <parser> ignores nested functions altogether, which is better. Candidate for documented limitation.
function out = function_scalar_return_nested( val ) out = 17; function nested end end
