Have the parser return some kind of data which can be easily traversed and apply callback functions on it.

~134ms parse with callbacks, ~66ms without, given the number of unused parses it might actually make a performance difference. Most unused parses never callback though so who knows.

No arguments are listed in their json file, presumably they are defined in the class WFContentFilterItemAction

Text
| hi
| ho
| hu

Error, could not convert. Took 55ms. undefined has no properties

Right now to edit a dictionary value

SetDictionaryValue ^v:Dictionary Key Value -> v:Dictionary

maybe some syntax

v:Dictionary <-> SetDictionaryValue Key Value

or similar

v:myvar{key:"hi"}

This should throw an error and require overwriting in the overrides file

For the @import action, generally this will happen

• The main file will be parsed and cached
• Using the caches, scripts will be parsed until an import action is called. If a cache exists for it, it will use the cache. (exits using throw probably, algebraic effects would be perfect for this but don't exist, a sync process otherwise needs to be frozen until an async finishes which is not possible in a browser, if using a warning it may take longer to do everything but then multiple files could be imported at once)
• The fileLoader function will be called on the file to be imported
• The loaded file will be parsed and cached
• The process repeats until all files are imported and converting finishes sucessfully

FormatDate's WFDateFormatStyle references an argument that isn't set yet in a relation resource...

Maybe do a manual override and detect and error on issues like these?

This syntax will be used in the inverse converter to signify that something went wrong. For better error messages, add support for it with a ErrorParse that throws this.error on all .as types

text ??huh there is an error??
getvariable s:??invalid variable??
text "My variable is: \(mv:??broken magic variable??)"

^( text ) is not valid and ^( text ) is also not valid, allow these.

This prevents undefined errors and stuff

Jest supports having tests next to normal files as filename.spec.ts . This would be useful to have in addition to tests that make sure scpl code compiles and inverses as expected.

use a.asAction(); cc.lastAction in places where a variable is needed. If cc.lastAction is undefined, throw an error saying an action is needed or something

Shortcuts 3.0 support will not be worked on until iOS allows shortcuts to be imported from .shortcut files. Until then, it is not known if it ever will and development cannot be tested.

Major Changes in Shortcuts 3.0:

Some actions, such as the CombineText action have been moved to separate files that includes all their information such as Name and Parameters (Presumably an intent declaration file but I haven't found it yet). Add support for loading action data from these files

Commonly used actions such as if now have their own custom parameter types and show different things based on the type of arguments passed into them. ScPL will now need to keep track of variable types at all times and use aggrandizements to make sure they always have as: set as ScPL assumes they do. (iOS might track variable as: differently). ScPL will also need significantly more information about types and have argument classes like for if with lots of different enum values based on other values.

Shortcuts now uses a natural language format for actions. ScPL could support this format. For example: Set variable (name) to (value). This might not be possible because of conflicts or might be a bad idea because you would constantly need to look up documentation.

Things that need to be changed in ScPL:

Change inputarg (^) to set the WFInput parameter instead of setting the action above.

For example:

in 2.2 that makes Text "hi" SetVariable v:myvar
in 3.0 it should make Text "hi" -> mv:var SetVariable input=mv:var value=v:myvar

Default actions that accept WFInput but don't have an inputarg parameter to use the action above them.

For example:

In 2.2, Text "hi"; SetVariable v works
In 3.0, it must be Text "hi" -> mv:var; SetVariable mv:var v

So, if an action does not have its WFInput defined, set it to the action above it.

Remove WFExpandingParameter, what's the point of it

Barlists should probably use their own chars definition rather than using the string chars

Text "test\ntest"

outputs "test\n\ntesttest"

Not sure how to fix this with the current parsing system.

Expected:

// Comment

compiles properly

Actual:

// Comment

throws a parser error with no explanation

something like undefinedaction, in shortcuts it would show as the action does not exist, in the wip web editor it would have a custom error render, cc could handle having multiple errors and this could allow having multiple errors.

Same thing for undefined magic variables and named variables, output an error and an invalid UUID to make it red.

The last action requires a newline, causing strange things like @{Text "hi"} to error. This was worked around before @{} by adding a newline to the end of the file but it needs a real fix now

For example,

Count Items -> (QuickLook)

|
v

Count Items -> mv:Items
QuickLook
GetVariable mv:Items

maybe, not sure about this one.

variables have asnametype and actions have asaction so check if asnametype (do variable things) else (at end, varflag.asAction(), action.variable)

Both yargs and chalk are only used for the CLI, but in some cases I am only using the Node.js API. Personally, I think it would be better to split this into a scpl-cli package and link between the two.

The highlights are removed but the error gutter marker is not

When importing shortcut https://www.icloud.com/shortcuts/f139ee120a2f44baace3fc01634408bd

The documentation link in the README leads to your main GitHub Pages site, which in turn redirects to pfg.pw, when it should lead to the subdomain.

ScPL only supports get:, as:, and key: right now. More such as those in the Date coercion type could be helpful.

Text "mytext" -> @:myvar should be allowed

Maybe sonething like

if Equals hi{
  do thing
}{
  don't
}

where each bracket increases the controlflow number

Actions have some data to say what their output type is. If cc keeps track of this, ScPL could guess what type a variable is and set it when using as:

For example, an error like variable does not exist could be shown in combination with other errors.

Also, allow some errors to generate output anyway. For example, magic variable does not exist could generate output with a broken magic variable (random unused uuid)

...

maybe bplistc is outputting -1 as an integer or something? and shortcuts reads it as an unsigned integer?

Maybe add support for a=b in arguments like

SetDictionaryValue key=hi value=bye

For example,

@set a b

@:a

would error at b (or both a and b)

@set c @:a

@:c{a:test}

would error at test rather than @:a

With filters now added, units can go in Data/GetTypes.ts

@seticon <shortcuts js icon name>
@setcolor <shortcuts js color name>

or something

shortcutsjs has all the icons

https://docs.shortcuts.fun/metadata/icon

It could be confusing that macros are defined no matter what happens in an if. This could be avoided if @set errors

Text "hi"
If Equals "something else"
  @set a b // always called no matter what
end

a{} is not great, maybe make it so barlists can have key=value or something

• Remove build from the git repo (Done in a6f037b)
• Parameters go in separate files (Done in da85d1c)
• Move addStaticEnums to OutActions generation
• Do something about WFAction
• Rename OutputData
• Move OutputData classes to their own files
• Rename ParserData and find a more descriptive word for *Parse
• Separate all *parses into their own separate files
• Make all files with real code have less than 100 lines (excluding data files like GetTypes)

@set a 5
@set b @:a
@set a 6
text @:b // 6

Is this a bug? Is it possible to make copies instead?

Error messages thrown by the converter have no location, making them hard to identify