balazsjako / imguicolortextedit Goto Github PK

How to change the size of the textbox

SetSelection(Coordinates(0, 0), Coordinates((int)mLines.size(), 0));
Would resolve to Coordinates(0,0), Coordinates(1,0), which is then sanitised by SetSelection to Coordinates(0,0),Coordinates(0,0), resulting in no selection as the end coord is beyond the end of the document.

I've fixed this locally by changing it to select the last character of the last line instead of the line beyond the end of the document:

void TextEditor::SelectAll()
{
	if(!mLines.empty())
		SetSelection(Coordinates(0, 0), Coordinates((int)mLines.size()-1, mLines.back().size()));
}

Return key does not work for me ...

I'm working on a tool, if you're interested: https://github.com/berdal84/Nodable/
I would like to be able to use your imGuiColorTextEdit with custom width/height fonts.
I'll be happy to contribute by adding custom font handling, are you open to it ? If yes, any suggestions to do it before I start ?

A capture of your ImGuiColorTextEdit integrated into my app :
https://github.com/berdal84/Nodable/blob/texteditor/screenshots/2018_06_03_Testing_ImGuiColorTextEdit.png

Origin: https://twitter.com/luksamuk/status/916003118575489024

So in my setup whenever I write _ (Shift + -) I get everything from my clipboard pasted in addition to the underscore character written. This is because of this code. What is the purpose behind pasting with _?

The ASCII code of - is 45 - is it supposed to be something else?

I also tested this with the demo project and it doesn't get triggered there - any idea why?

I'm using this code from the examples of imgui for opengl2 and glfw

Colorizer uses std::regex, but it is too slow for that amount of work. Not to mention it is not optimized to scan multiple expressions at once. It is necessary to replace it with a proper lexical scanner (like RE/Flex).

Problem

When I click on a line contening tabulations, the cursor is placed at a wrong position.

For example, calling only these two functions with the default font of ImGui:

editor.SetText("        abcdefghijklmnopqrstuvwxyz\n\t\tabcdefghijklmnopqrstuvwxyz\n");
editor.Render("editor", ImGui::GetContentRegionAvail(), false);

If I click after the text on the first line the cursor is put at the right position (the end of the line), but if I click after the end of the second line the cursor is placed at a wrong position (as in the picture).

Informations

• ImGui version: v1.70+, docking branch (ocornut/imgui@596d81a)
• ImGuiColorTextEdit version: last (e2a2520)

https://imgur.com/a/tFBKN

Its cool that we can go through words with Ctrl+ Left/Right and etc. but perhaps more standard controls like Ctrl+A for selecting everything (and others I can't think of right now) might be a big plus :)

Removing a line using delete removes the marker on that line, as expected.
Removing a line using backspace however, doesn't. Using backspace to delete the line below the marker does remove it though.

A bit of an inconsistency in their behavior.

I will try to make a fix myself. But that requires rewriting either Delete() or Backspace(), while maintaining the other existing behavior.

Longer the line more cursor gets off to right from place it is writing. I tried multiple fonts along with monospaced one but nothing helped. Maybe some change in recent ImGui? I'am using release of 1.68 made recently. Here is the issue visible:

Cursor is there actually at the end of the line (when i press delete last character on line dissapears)

It should return the std::string inside the object instead of a copy. The string might destroyed when the function exits.

Example.

Most languages have completely different tokens for single-line comments and block comments (eg // and /* */ in C), but in Lua they start the same: -- and --[[.
The comment highlighting logic needs fixing so that it prioritises block comments first.

I've fixed this locally by swapping the logic over:

if(currentIndex + startStr.size() <= line.size() &&
	equals(startStr.begin(), startStr.end(), from, from + startStr.size(), pred))
{
	commentStartLine = currentLine;
	commentStartIndex = currentIndex;
}
else if (singleStartStr.size() > 0 &&
	currentIndex + singleStartStr.size() <= line.size() &&
	equals(singleStartStr.begin(), singleStartStr.end(), from, from + singleStartStr.size(), pred))
{
	withinSingleLineComment = true;
}

If I set an empty string :
editor.SetString("");
And then I click into the Text Editor area I got an assert (aLine is wrong).

Also it would be grand if the double-click glob chars could be changed, I'm thinking about ":" specifically. In passwd and shadow files It's great to stop at : but for "scp host:file ." it's unfortunate.

There are two issues introduced in 8fbb932:

• langDef.mSingleLineComment = "--"; was removed from the Lua language definition.
• ColorizeInternal() will set withinSingleLineComment to true if mSingleLineComment is the empty string, so a language that does not have comments (e.g. JSON) will be fully colorised as comments

g++ -I../cimgui/imgui -I../../include   -c -o TextEditor.o TextEditor.cpp
In file included from TextEditor.cpp:7:0:
TextEditor.h:125:30: warning: ‘TextEditor::Glyph::mColorIndex’ is too small to hold all values of ‘enum class TextEditor::PaletteIndex’
   PaletteIndex mColorIndex : 7;

This is on i686, 32bit linux.

I'm basically using similar logic, but I'm not sure whether I need the resize there at the end.
Any advice on why you have it there?

Notice how the red marking goes to the left when I scroll to the right

1. Awesome work!

2. PTAL at https://github.com/ocornut/imgui_club/tree/master/imgui_freetype
   It has hinting support so it will result in better text render quality.
   (you'll find screenshots here: ocornut/imgui#618)

We've noticed that there's a lag colorizing new text - as you're typing it renders in the default color for a frame before the syntax highlighting kicks in.

It appears this is happening because when ColorizeInternal goes down the multi-line comment path, it returns rather than continuing to do the main colorization pass. This results in a two-stage process where commenting is checked on frame 1, and then syntax highlighting applies on frame 2.

Is there a reason for it being this way? I can see any bad effects after removing the return locally. According to blame, it's always been there.

When a code editor widget is focused and has a cursor inside of it - querying ImGui::GetIO().WantCaptureKeyboard or ImGui::GetIO().WantTextInput (or maybe both) should return true instead of false. Also not sure about WantCaptureMouse and the rest such flags.

Great project btw! Thanks for making it public!

The following warning is produced from the header with gcc 7:

TextEditor.h:125:30: warning: 'TextEditor::Glyph::mColorIndex' is too small to hold all values of 'enum class TextEditor::PaletteIndex'
   PaletteIndex mColorIndex : 7;

on the default warning level - and it cannot be easily silenced (without disabling all warnings with -w)

Should reflect the character underneath.

You can see some examples here http://lua-users.org/wiki/StringsTutorial

The way you currently check for multiline strings and comments won't allow for the nesting capability that lua has.

Minor trick/suggestion: because png adds up in the git repo size.
The wiki that github offers is a separate git repo (ImGuiColorTextEdit.wiki.git instead of ImGuiColorTextEdit.git)

What I do for imgui is I upload any images to the wiki repo and link to them from the readme instead, so the main repository doesn't grow with extra data.

Inside GetDarkPalette(), GetLightPalette() and GetRetroBluePalette()

error: suggest braces around initialization of subobject [-Werror,-Wmissing-braces]

I've fixed this locally by double-bracing the initialisation lists so they look like this:

static Palette p = {{ 
		0xffffffff,	// None
		0xffd69c56,	// Keyword	
		0xff00ff00,	// Number
		0xff7070e0,	// String
		0xff70a0e0, // Char literal
		0xffffffff, // Punctuation
		0xff409090,	// Preprocessor
		0xffaaaaaa, // Identifier
		0xff9bc64d, // Known identifier
		0xffc040a0, // Preproc identifier
		0xff206020, // Comment (single line)
		0xff406020, // Comment (multi line)
		0xff101010, // Background
		0xffe0e0e0, // Cursor
		0x80a06020, // Selection
		0x800020ff, // ErrorMarker
		0x40f08000, // Breakpoint
		0xff707000, // Line number
		0x40000000, // Current line fill
		0x40808080, // Current line fill (inactive)
		0x40a0a0a0, // Current line edge
	}};

It would be nice if the rendered text could wrap around (optionally with some flag) so we don't have to scroll horizontally when the lines are too long

I understand that this project doesn't aim to be a complete editor - I'm just suggesting... :D

how to add the text edditor to imgui

"foo\"bar", the \ should be highlighted... IMHO.

Hi,

The following code has incorrect syntax highlighting. One can verify this by copy-pasting it into the editor.

	if (hasValue == false && nodeId != kGraphNodeIdInvalid)
	{
		//logDebug("reset realTimeSocketCapture");
		//*this = GraphEdit_Visualizer();
	}
	
	//
	
	const bool hasChannels = channelData.hasChannels();
	
	if (hasChannels)
	{

I investigated why, and it's caused by the multi-line comment highlighter running before single line comments are detected. It causes the '/*' in '//*this' above to be interpreted as the start of a multi-line comment. While it should be ignored as it's part of a single line comment. A possible solution would be to detect single line comments together with multi-line ones.

Perhaps eventually to make (multi-line) comments work well across different languages we'll need a separate comment highlighter for each language.

with GCC the <cmath> header needs to be included for floor() in the .cpp - this PR fixes that.

Currently I have a workaround in a demo project (as can be seen here) which I would like to remove so that the code is simpler - people will be looking at it when I make an official release of that example project.

Hi, I hope you can read that :). I encountered a little problem with fonts, it looks like you use the first one in the array of fonts in ImGui, and if you are using different fonts that doesn't seem to work...

void TextEditor::Render(const char* aTitle, const ImVec2& aSize, bool aBorder)
{
	mWithinRender = true;

	ImGuiIO& io = ImGui::GetIO();
	auto xadv = (io.Fonts->Fonts[0]->IndexAdvanceX['X']);
	mCharAdvance = ImVec2(xadv, io.Fonts->Fonts[0]->FontSize + mLineSpacing);

My solution was passing the used font as an argument in the "Render" function, although I think that it'd be better to acces the currently used font from the ImGui io.

void TextEditor::Render(const char* aTitle,ImFont* font, const ImVec2& aSize, bool aBorder)
{
	mWithinRender = true;

	ImGuiIO& io = ImGui::GetIO();
	auto xadv = (font->IndexAdvanceX['X']);
	mCharAdvance = ImVec2(xadv, font->FontSize + mLineSpacing);

Having said that, your API i great, the fact that you can make custom colors for different languages is really good. I'll make one for "wren", and pass it to you in case you want to merge it with the source code.

There are a few [ \t] that should be [ \\t].

I can't wait to combine this with http://duktape.org and truly empower my users.

Hi! currently the source code requires c++14.
Only 2 lines of code have to be changed to enable c++11 builds - where std::equals() is used.

Including this lib in my project causes the time from starting the executable to the time the window is displayed to be increased by several seconds.

I would appreciate any input about this issue, such as why it might be so long, and what if anything I might do about it.

I really like this lib otherwise, but more than tripling my startup time is hard to accept.

FWIW I'm using this with ImGui 1.69, sdl2 and gl3w. I'm using the default impl_sdl and impl_opengl3.

in:
void TextEditor::HandleKeyboardInputs()

this row:
else if (!IsReadOnly() && !ctrl && !alt)

makes the texteditor unable to enter any characters requiring AltGr key to be pressed (coded as Ctrl+Alt in European Windows). (for instance { } [ ] etc)

Quick fix for those affected is to remove the !ctrl and !alt check.

This issue doesn't exist within ImGui's inputtext itself.

Do you reckon there's a better fix for this issue?

Would be nice if there was an easy way to make the editor read only. Selections and such would still be possible, but text input, deletions etc would be locked.

It may be possible now. Maybe I just missed it. If not, it would be nice.

My project builds and runs, but the enter key does not feed a line and return the carrage.

I've installed the callbacks and passed on keypresses.

This is a nice addon. I got a question when using it. Should io.FontGlobalScale be counted here:

Thanks.

Eg. BASIC, etc.

What can I do if I need this Text Editor in c#?

The readme says no UTF8 support... how far would that be on the horizon? Would it involve a lot of changes?

1. Select random code (which is indented tab)
2. Shift + tab
3. Tab
4. Ctrl + z
5. Crash

fix ?

Probably needs a languagehelper-like class.

It would be useful to be able to receive information about when the text has been changed (eg to trigger a Lua script load to check for errors in realtime).

Naively, this could be done by the user comparing GetText() with the previous value, but it's a bit of a sledgehammer approach.

Better would be implemented with a "text changed" member bool that's set inside any text manipulation function that adjusts the undo index. The flag would be returned and cleared at the end of Render().