Fast path for handling fixed length strings?

[Andersama]

Was looking at text rendering code and I spotted that "%s" is specifically tested likely as a fast path for formatting text.

void ImGui::TextColoredV(const ImVec4& col, const char* fmt, va_list args)
{
    PushStyleColor(ImGuiCol_Text, col);
    if (fmt[0] == '%' && fmt[1] == 's' && fmt[2] == 0)
        TextEx(va_arg(args, const char*), NULL, ImGuiTextFlags_NoWidthForLargeClippedText); // Skip formatting
    else
        TextV(fmt, args);
    PopStyleColor();
}

Was wondering if it may be worth having a fast path for the "%.*s" format. If I remember correctly the format uses an int argument first followed by a char pointer, so not that much different from "%s". I don't use c variadics much, I'm assuming something like the following works:

void ImGui::TextColoredV(const ImVec4& col, const char* fmt, va_list args)
{
    PushStyleColor(ImGuiCol_Text, col);
    if (fmt[0] == '%' && fmt[1] == 's' && fmt[2] == 0)
        TextEx(va_arg(args, const char*), NULL, ImGuiTextFlags_NoWidthForLargeClippedText); // Skip formatting
    else if (fmt[0] == '%' && fmt[1] == '.' && fmt[2] == '*' && fmt[3] == 's' && fmt[4] == 0)
       TextEx(va_arg(args, int), va_arg(args, const char*), ImGuiTextFlags_NoWidthForLargeClippedText); // Skip formatting
    else
        TextV(fmt, args);
    PopStyleColor();
}

[ocornut]

It's possible indeed and may be worthy as we transition to string view branch.
Curious if you actually need/use it or if you suggested this without a strong need?

[Andersama]

It was a passing thought, actually while I was looking at this earlier is there any concern about accessing memory out of bounds with just const char* fmt?

Presumably like with other functions the format string would also be a good thing to know the length of.

Working on a text editor as a personal project.

[PathogenDavid]

is there any concern about accessing memory out of bounds

Since fmt is null-terminated it should always be at least one char.

For an empty string (IE: fmt = { '\0' }), the fmt[0] == '%' condition will be in-bounds and result in false, which will short-circuit/skip the other checks (which would've been out-of-bounds.)

[Andersama]

I suppose my gut instinct is because there's a missing null check that something would go wrong, but that's a different out of bounds condition than the one I was thinking of, so I almost feel silly asking.

[PathogenDavid]

No worries, we all get brain farts.

Dear ImGui is old school and generally doesn't explicitly check for basic mistakes like passing null for something that should never be null.

[ocornut]

Pushed this change with d486920
Thanksfully the code has been massaged (you are on an older version) so there's only 1 spot to modify now.
I don't feel this is very important but as we'll aim to better support languages which might encourage users to perform formatting on their side, this seems to make sense going ahead.