SynEdit Highlighter

From Lazarus wiki
Revision as of 20:40, 28 April 2015 by Life4YourGames (talk | contribs)
Jump to navigationJump to search

For more info on SynEdit go to: SynEdit


Understanding the SynEdit Highlighter

SynEdit - Highlighter relationship

SynEdit <-> Highlighter have an n to 1 relationship.

  • 1 (instance of a) Highlighter can serve n (many) (instances of) SynEdits
  • Each SynEdit only has one Highlighter
  • But: one text (text-buffer) can have many highlighters, if shared by several SynEdit (each SynEdit will have one HL, but all HL will work on the same document)

As a result of this:

  • no Highlighter Instance has a (fixed) reference to the SynEdit.
(Highlighters however keep a list of SynEditTextBuffers to which they are attached)
  • All data for the Highlighter is (and must be) stored on the SynEdit (actually on the TextBuffer of SynEdit (referred to as "Lines").

However, before each call to the Highlighter, SynEdit ensures that Highlighter.CurrentLines is set to the current SynEdits Lines. This way the highlighter can access the data whenever needed. The Format of the data storage is determined by the highlighter (TSynCustomHighlighter.AttachToLines).

Scanning and Returning Highlight attributes

The Highlighter is expected to work on a per Line base.

If any text was modified, SynEdit will call (TSynCustomHighlighter.ScanFrom / Currently called from TSynEdit.ScanFrom) with the line range. The Highlighter should know the state of the previous line.

If Highlight attributes are required, SynEdit will request them per Line too. SynEdit will loop through individual tokens on a line. This currently happens from nested proc PaintLines in SynEdit.PaintTextLines. It calls TSynCustomHighlighter.StartAtLineIndex, followed by HL.GetTokenEx/HL.GetTokenAttribute for as long as HL.GetEol is false.

Also the BaseClass for the Highlighter's data (see AttachToLines) is based on per line storage, and SynEdit's TextBuffer (Lines) do maintenance on this data to keep it synchronized. That is: whenever lines of text are inserted or removed, so are entries inserted or removed from the highlighters data (hence it must have one entry per line).

Usually Highlighters store the end-of-line-status in this field. So if the highlighter is going to work on a line, it will continue with the state-entry from the previous line.

Folding

SynEdit's folding is handled by unit SynEditFoldedView and SynGutterCodeFolding. Highlighters that implement folding are to be based on TSynCustomFoldHighlighter.

The basic information for communication between SynEditFoldedView and the HL requires 2 values stored for each line. (Of course the highlighter itself can store more information):

  • FoldLevel at the end of line
  • Minimum FoldLevel encountered anywhere on the line

The Foldlevel indicates how many (nested) folds exist. It goes up whenever a fold begins, and down when a fold ends:

                           EndLvl   MinLvl
 Procedure a;               1 -      0
 Begin                      2 --     1 -
   b:= 1;                   2 --     2 --
   if c > b then begin      3 ---    2 --
     c:=b;                  3 ---    3 ---
   end else begin           3 ---    2 --
     b:=c;                  3 ---    3 ---
   end;                     2 --     2 --
 end;                       0        0  // The end closes both: begin and procedure fold

In the line

 Procedure a;               1 -      0

the MinLvl is 0, because the line started with a Level of 0 (and it never went down / no folds closed). Similar in all lines where there is only an opening fold keyword ("begin").

But the line

   end else begin           3 ---    2 --

starts with a level of 3, and also ends with it (one close, one open). But since it went down first, the minimum level encountered anywhere on the line is 2.

Without the MinLvl it would not be possible to tell that a fold ends in this line.

There is no such thing as a MaxLvl, because folds that start and end on the same line can not be folded anyway. No need to detect them.

 if a then begin b:=1; c:=2; end; // no fold on that line


Creating a SynEdit Highlighter

Since 0.9.31 Revision 35115 the fold-highlighter has changed. Implementing basic folding is now easier.

All Sources can be found in the Lazarus installation directory under:

 examples\SynEdit\NewHighlighterTutorial\

The project HighlighterTutorial contains 3 difference example highlighters:

  • SimpleHl: used in Step 1 below
  • ContextHl: used in Step 2 below
  • FoldHl: used in Step 3 below

SimpleHl and ContextHl will work with 0.9.30 too

The folding highlighter in action: SynEditFoldingHighlighterDemo.png

The Basics: Returning Tokens and Attributes

As indicated, the SimpleHl unit demonstrates this process.

What it does

  • It splits each line into words and spaces (or tabs)
    • The spaces are part of the text, and must be highlighted too
  • This example allows specifying different colors for
- text (defaults to not-highlighted)
- spaces (defaults to silver frame)
- words, separated by spaces, that start with a,e,i,o,u (defaults to bold)
- the word "not" (defaults to red background)

How it works

  • Creation

The Highlighter creates Attributes that it can return the Words and Spaces.

  • SetLine

Is called by SynEdit before a line gets painted (or before highlight info is needed)

  • GetTokenEx, GetTokenAttribute, Next, GetEol

Are used by SynEdit to iterate over the Line. Note that the first Token (Word or Spaces) must be ready after SetLine, without a call to Next.

Important: The tokens returned for each line must represent the original line-text, and be returned in the correct order.

  • GetToken, GetTokenPos, GetTokenKind

SynEdit uses them e.g for finding matching brackets. If tokenKind returns different values per Attribute, then brackets only match, if they are of the same kind (e.g. if there was a string attribute, brackets outside a string would not match brackets inside a string).

Other notes

For readability, the highlighter has no optimization, so it may be very slow on larger texts. Many of the supplied highlighters use hash functions, to find what word (or any group of chars) is.

Step 2: Using Ranges

As indicated, the ContextHl unit demonstrates this process

The next example allows content of a line that influences other lines that follow. An example: a "(*" in Pascal makes all following lines a comment until a "*)" is found.

This example extends the SimpleHl show above: The tokens -- and ++ (must be surrounded by space or line-begin/end to be a token of their own) will toggle words that start with a,e,i,o,u

Multiple ++ and -- can be nested. Then for each -- a ++ must be given, before the words highlight again.

Then we extend the scanner. The pre-scan to store the information calls the same functions as the highlighter. It is automatically called, if anything changes. (It is called for all lines below the changed line, until a line returns the same Range-value as it already had)

The current amount of "--" is counted in

  FCurRange: Integer;

The amount is decreased by "++"

To store the info we use:

GetRange
Called after a line is completely scanned, to get the value at the end of the line. The value will be stored.
SetRange
Called before a line gets scanned. Sets the value stored from the end of the previous line.
ResetRange
Called before the 1st line is scanned (as there is no previous line).
Light bulb  Note: A scan is triggered by *every* change to a line (every keystroke). It scans the current line, and all lines below, until a line returns the same range that it already had. See: http://forum.lazarus.freepascal.org/index.php/topic,21727.msg139420.html#msg139420

Step 3: Add Folding

As indicated, the FoldHl unit demonstrates this process

For the example, the highlighter should fold everything between free-standing "-(-", "-)-".

Change inheritance:

  uses SynEditHighlighterFoldBase;
  ...
  TSynDemoHl = class(TSynCustomFoldHighlighter)

Change the way range info is stored, since the base class uses it for fold-info:

procedure TSynDemoHl.SetRange(Value: Pointer);
begin
  inherited;
  FCurRange := PtrInt(CodeFoldRange.RangeType);
end;

procedure TSynDemoHl.ResetRange;
begin
  inherited;
  FCurRange := 0;
end;

function TSynDemoHl.GetRange: Pointer;
begin
  CodeFoldRange.RangeType := Pointer(PtrInt(FCurRange));
  inherited;
end;

Now add code to the scanner which tells the highlighter about opening and closing folds:

procedure TSynDemoHl.FindTokenEnd;
begin
   ...

  if (FTokenEnd = FTokenPos+1) and (FLineText[FTokenPos] = '[') then
    StartCodeFoldBlock(nil);
  if (FTokenEnd = FTokenPos+1) and (FLineText[FTokenPos] = ']') then
    EndCodeFoldBlock();
end;
  • For 0.9.30

Please see the history of this page, if you are using a 0.9.30 version [[1]]

More info on StartCodeFoldBlock / EndCodeFoldBlock

  function StartCodeFoldBlock(ABlockType: Pointer; IncreaseLevel: Boolean = true): TSynCustomCodeFoldBlock; virtual;
ABlockType
Can be used to specify an ID for the block.

The field is not normally used as a pointer (though this is permitted). Normally IDs are a numeric enumeration.
If you have different types of block (e.g. in Pascal: begin/end; repeat/until, ...), and you do not want a block being closed by the wrong keyword ("end" should not close "repeat"), then you can give them IDs:
StartCodeFoldBlock(PtrUInt(1)) // or other numbers

IncreaseLevel
If set to False, then a block will be inserted that can not be folded.
Blocks like that can be used for internal tracking.

NOTE: All folds must be nested, they can not overlap. That is, the last opened fold must be closed first.
This refers to the "EndLvl" as shown in "Folding"(1.3) above
Overlaps (like IFDEF and begin in the IDE) can not be done this way.

  procedure EndCodeFoldBlock(DecreaseLevel: Boolean = True); virtual;
DecreaseLevel
This *must* match IncreaseLevel, as it was given on the StartCodeFoldBlock

True means a fold ends; False means an internal block is ended. If mismatched then folds will either continue, or end early.
TopCodeFoldBlockType can be used to indicate the ID of the innermost open block. One can use different IDs for internal blocks, and use this to set the value.

  function TopCodeFoldBlockType(DownIndex: Integer = 0): Pointer;

Returns the ID of the innermost block.

DownIndex
can be used to get ID for the other block.

DownIndex=1 means the block that is surrounding the innermost.

Edit an existing highlighter (Example for pascal)

Light bulb  Note: This example is done with the IDE "lazarus", your IDE might behave or be different in some aspects up to not using SynEdit for its Syntax-Highlighting at all

Sometimes you might want to edit existing highlighters (just like I wanted a few days ago) that already exist. In this example we're going to edit the highlighter for pascal-like code (classname: TSynPasSyn; package: SynEdit V1.0; unit: SynHighlighterPas.pas).

Say, what we want to reach is, that our application (lazarus in this case) differs between the three types of comments, which exist in Pascal:

  (* ansi *)
  { bor }
  // Slash

This may be helpful, if you want to differ between different types of your comments (e.g. "Description", "Note", "Reference", etc.) and want them to be e.g. colored in different ways.

How to:

Light bulb  Note: Just in case you break something, I suggest to make some "NEW" and "/NEW"-comments, but you don't have to
  • First, open the unit "SynHighlighterPas" which should be located in your SynEdit-directory.
  • As we don't want to cause incompatibilities, we're creating a new type of enumerator which helps us to identify our comment later:

E.g. under the declaration of "tkTokenKind", write this:

  {NEW}
  TtckCommentKind = (tckAnsi, tckBor, tckSlash);
  {/NEW}
  • In the declaration of "TSynPasSyn" search for "FTokenID" and add the following between "FTokenID" and the next field
  {NEW}
  FCommentID: TtckCommentKind;
  {/NEW}
  //This creates a new field, where we can store the information, what kind of comment we have
  • In the declaration of "TSynPasSyn" search for "fCommentAttri" and add the following between "fCommentAttri" and the next field
  {NEW}
  fCommentAttri_Ansi: TSynHighlighterAttributes;
  fCommentAttri_Bor: TSynHighlighterAttributes;
  fCommentAttri_Slash: TSynHighlighterAttributes;
  {/NEW}
  //This allows us, to return different Attributes, per type of comment
  • Next, search for the constructor-definition of "TSynPasSyn", which should be "constructor TSynPasSyn.Create(AOwner: TComponent);"
  • We need to Create our new Attributes, thus we add our Attributes somewhere in the constructor (I suggest, after the default "fCommentAttri")
  (...)
  AddAttribute(fCommentAttri);
  {NEW}
  fCommentAttri_Ansi := TSynHighlighterAttributes.Create(SYNS_AttrComment+'_Ansi', SYNS_XML_AttrComment+'_Ansi'); //The last two strings are the Caption and the stored name
  //If you want to have default settings for your attribute, you can e.g. add this:
  //fCommentAttri_Ansi.Background := clBlack; //Would set "Background" to "clBlack" as default
  AddAttribute(fCommentAttri_Ansi);
  fCommentAttri_Bor := TSynHighlighterAttributes.Create(SYNS_AttrComment+'_Bor', SYNS_XML_AttrComment+'_Bor');
  AddAttribute(fCommentAttri_Bor);
  fCommentAttri_Slash := TSynHighlighterAttributes.Create(SYNS_AttrComment+'_Slash', SYNS_XML_AttrComment+'_Slash');
  AddAttribute(fCommentAttri_Slash);
  {/NEW}
  (...)
  • The "complex" part now is, to search for the points, where "FTokenID" is set to "tkComment" and to set our "subtype", equally (of course, I've already searched them:)
procedure TSynPasSyn.BorProc;
(...)
  fTokenID := tkComment;
  {NEW}
  FCommentID:=tckBor;
  {/NEW}
  if rsIDEDirective in fRange then
(...)
procedure TSynPasSyn.AnsiProc;
begin
  fTokenID := tkComment;
  {NEW}
  FCommentID:=tckAnsi;
  {/NEW}
(...)
procedure TSynPasSyn.RoundOpenProc;
(...)
        fTokenID := tkComment;
        {NEW}
        FCommentID:=tckAnsi;
        {/NEW}
        fStringLen := 2; // length of "(*"
(...)
procedure TSynPasSyn.SlashProc;
begin
  if fLine[Run+1] = '/' then begin
    fTokenID := tkComment;
    {NEW}
    FCommentID:=tckSlash;
    {/NEW}
    if FAtLineStart then begin
(...)
procedure TSynPasSyn.SlashContinueProc;
(...)
    fTokenID := tkComment;
    {NEW}
    FCommentID:=tckSlash;
    {/NEW}
    while not(fLine[Run] in [#0, #10, #13]) do
(...)
  • Now, we just have to retreve the information when "GetTokenAttribute" is called and return the right Attribute, therefore we edit "GetTokenAttribute" as follows:
function TSynPasSyn.GetTokenAttribute: TSynHighlighterAttributes;
begin
  case GetTokenID of
    tkAsm: Result := fAsmAttri;
    {OLD
    tkComment: Result := fCommentAttri; //This is commented and just backup, so it'll be ignored
    /OLD}
    {NEW}
    tkComment: begin
      if (FCommentID=tckAnsi) then Result:=fCommentAttri_Ansi //Type is AnsiComment
      else
      if (FCommentID=tckBor) then Result:=fCommentAttri_Bor //Type is BorComment
      else
      if (FCommentID=tckSlash) then Result:=fCommentAttri_Slash //Type is SlashComment
      else
        Result:=fCommentAttri //If our code failed somehow, fallback to default
    end;
    {/NEW}
    tkIDEDirective: begin
(...)

If you do use lazarus, just reinstall the SynEdit-Package, if not, recompile your project/the package/<similar>.

DONE ! No seriously, you are now ready to differ between the different types of comments.

Aditional Info

The lazarus-IDE does automatically detect, what attributes exist and shows them in the options, such as saves them, if you change them. If your application/IDE doesn't do this, you will have to set Color/Font/etc. of the new Attributes somewhere manually (e.g. in the constructor of TSynPasSyn)

--Life4YourGames 20:40, 28 April 2015 (CEST)


References

Threads on the forum:

  • Folding
    • "SynEdit - improved highlighters to handle Code Folding?" [topic,7879]
    • "SynEdit - Add support code folding for Java" [topic,7338]
    • "CodeFolding" Config [topic,11064]
    • Fold blocks "end" keyword (end on the line before the next keyword) [topic,23411.msg139621]
    • Folding selected text from code (user/application code): [24473.msg147312]
    • Obtaining state of folding (save fold state with session) [topic=26748]