derula / strainer Goto Github PK

• Save existing settings using QSettings
• Use saveState and saveGeometryto save / restore user settings

A new lexer and parser are in development, They are currently much slower than the sievelib parser, but should also be more accurate / resistant to errors. Plus, since they are based on a grammar, they are more easily extendable / bugs should be easier to fix.

To do:

• Rewrite Scintilla lexer to use new lexer
• Run new parser to find errors
  • Make sure Parser only runs some time after typing stopped, because it is so slow
  • Consider finding all parse errors at once and adding a panel to show all of them => #31
• Consider adding more features to the grammar, e.g. Sieve extensions => #32

See also #16.

Currently, partial lexing uses a strange logic of "moving the goalposts" before lexing starts. This, however, potentially doesn't perform well, and causes other problems (cf. #15).

Instead:

• Send complete document to lexer
• "Fast forward" while the end of the token is before the start of the range
• Style all following tokens
• Stop styling once the start of the token is after the end of the range

1. Show full parse results using sievelib.
2. Show reason provided by server if upload fails due to parse error.

Lark has an option to give an error callback instead of raising on first error. With this, we could add a list of all errors instead of just highlighting the first.

The sieve documentation window currently has no option to return to the main page, making it difficult to look at other parts of the documentation unreachable by clicking on things. A button could either be added to a common (cf. #2) or separate toolbar (with actions loaded from the QWebEngine context menu), or just as a new entry in the context menu.

Currently, scripts can be opened and edited, but not saved. We need:

• Either a save button (toolbar? regular button?) or some kind of menu.
• Indicator of whether the script has been changed.
• Warning when closing with unsaved changed.

Currently it takes a long time before the window is shown, making it seem like something's not right.

Possible solutions:

• Show a splash screen (meeh)
• Lazy loading of settings / populating accounts list (only a partial solution, and a bit of a challenge)

Probably not random, but I couldn't reproduce them well yet.

Dearly needed.

There is already GUI in place to activate / delete scripts, it's just not functional yet. Should be done after #2 to avoid accidentally deleting scripts while testing and then not having any test data left.

When the status bar content is set to a long string, the minimum width of the window is adjusted to show it in full. This should not happen; status bar text should just be limited to the available width.

Currently, we rely on the sievelib lexer to style the script in Scintilla. Unfortunately, this lexer returns multi-line comments and strings as a single token, which breaks the ability to use Scintilla's partial restyling. (Neither "text:\n" not "/* comment\n" are recognized as the beginning of a string / comment in isolation.)

Possible solutions:

• Override the rules sent to the lexer (defeats the purpose of using a lib)
• Implement in sievelib and issue a PR (unclear if or when it will be merged)
• Don't rely on sievelib for lexing at all
• Always lex the entire file (current "solution")

For renaming scripts, the save button is grayed out if the name already exists. Accounts should do the same.

Leads to page not found. Should open the page for if.

• Sieve extensions?
• Explicitly recognize the optional hash comment in multi-line text?
• Understand ${hex:...} / ${unicode:...} syntax?

This is currently a hacky re-use of the Scintilla default menu and can't be used in the toolbar. Should reimplement all fioelds and add to toolbar.

It works... okay, most of the time. Sometimes it's stuttering. But if you have 2 accounts set up and restart the program, it completely crashes without even so much as an error message. It's likely that the way it's currently done is really stupid and based on a misunderstanding of how Qt works.