propensive / escapade Goto Github PK

If a string contains a URL link, its length is miscalculated. This may require an additional field to hold the link.

Escapes such as \n and \t are treated as literals inside ansi"" interpolated strings. The workaround is to embed a character (e.g. ansi"${'\n'}", but it's a lot of ceremony and is easily forgotten. There's code in Gossamer to expand standard escaped characters.

It's too confusing having both output and display/Displayable terminology.

Escapes are currently represented as a pair of objects, a "push" and a "pop", which add and then remove a style from the text. These are stored independently, which does not make so much sense as storing each pair as a single "range" object.

Amongst other problems, the documentation still refers to the interpolator as ansi"" rather than e"".

The way ANSI strings are rendered is not tested, except by looking at them to see if they seem right. It seems to work, but complex combinations of escapes are rarely used in practice, and any tests usually require a user to visually check that the text is rendered correctly in a terminal (and also to know what the correct rendering should look like).

A zero-length range, e.g. out"$Red(${t""})" should not emit any escape codes, since doing so may cause unreliable rendering. In particular, they can clobber subsequent escapes.

The $TTY environment variable should contain enough information to determine whether the current terminal supports all or any Escapade's output features, notably color. This environment could be provided through a context value in print methods.

Escapade assumes that every character is rendered as a single "cell" in the terminal, however zero-width spaces and double-width characters may deviate from this assumption, and there's currently no way to know the rendered width, even though this would be useful for Escritoire.

Standard Java functions don't seem to exist for determining the render width from a character (and it might be font-dependent and terminal dependant—this should be checked). One possible way to work it out would be to attempt to render each and every character in the terminal, checking the cursor position before and after rendering. This "profile" could then be saved... but more likely it's standard anyway.

This should be used, if available, when converting from a Show[T] to an AnsiShow[T] to apply some style to the text.

The combination of foreground/background color, bold, underline, italic, etc, could all be represented with a single Long which would be much faster than the current implementation using object copying.