Refine definition of extra metadata tags

[samuelallan72]

Fix potential ambiguities or issues with extra metadata:

1. original rules disallowed colons in value. This is unnecessarily restrictive: everything is whitespace separated, so allowing colons in the value of extra metadata does not introduce any ambiguity in parsing.

This allows extra metadata like key:val:ue to be correctly parsed as the key key and value val:ue.

2. Make it clear than the value may be empty. This was previously not defined. Allowing an empty value is also syntactically unambiguous, so this should be fine.

This allows an extra metadata entry like key:, where the key is key and the value is an empty string.

3. Make it clear that the key must not be empty. This was previously not defined. It does not make sense for there to be an extra metadata where the key is an empty string. As such, entries such as : or :something must not be parsed as extra metadata tags (but rather as part of the description text).

[clach04]

@swalladge nice PR, really great description in the PR for the "why" and implications/examples.

Fix potential ambiguities or issues with extra metadata:

1. original rules disallowed colons in `value`.
   This is unnecessarily restrictive:
   everything is whitespace separated,
   so allowing colons in the value of extra metadata does not introduce
   any ambiguity in parsing.

This allows extra metadata like key:val:ue to be correctly parsed as the key key and value val:ue.

👍 I like it, this explanation would be a good addition to the readme

2. Make it clear than the value may be empty.
   This was previously not defined.
   Allowing an empty value is also syntactically unambiguous,
   so this should be fine.

This allows an extra metadata entry like key:, where the key is key and the value is an empty string.

I'm less keen on this, consider an entry such as:

play game Phoenix Wright: Ace Attorney - extra details https://en.wikipedia.org/wiki/Phoenix_Wright:_Ace_Attorney

This would then end up with a key of "Wright" which I would argue is not useful. Colons are used in English in this form so this could leave to a number of these empty ones (we already have the issue with URLs but that's already part of the spec)

3. Make it clear that the key must not be empty.
   This was previously not defined.
   It does not make sense for there to be an extra metadata
   where the key is an empty string.
   As such, entries such as `:` or `:something`
   must not be parsed as extra metadata tags (but rather as part of the description text).

Agreed, again I think what you wrote for the PR is a good additional explanation to add to the readme :-)

Also see https://github.com/todotxt/todo.txt/issues/77 which I think is relevant.

I'm hoping we get more reviewer comments on this as all of the above is my interpretation.

Thanks for taking the time to think about this and contribute.

[clach04]

@xsrvmy and @jgclark thoughts?

[samuelallan72]

Thanks for the feedback @clach04 !

Re the empty value part - what is the current assumption, considering that it wasn't previously defined? (For what it's worth, I'm not fussed either way, as long is it's clearly defined. In my workflows, I've used an empty value as a 'null', but it could just as easily be a literal 'due:null' for example.)

Colons are used in English in this form so this could leave to a number of these empty ones (we already have the issue with URLs but that's already part of the spec)

Also see https://github.com/todotxt/todo.txt/issues/77 which I think is relevant.

So for these examples (including #77), I would argue that this is a separate issue to the syntax definition. This is more of a language or domain specific data issue, which I don't think we can or should fix through syntax rules. Many things have embedded colons, with or without whitespace: times (16:35), mac addresses (4d:34:c2:...), sentences (example: this is an example), urls (ftp://server, http://..., https://..., file://..., file:name), colon separated data (column1:column2:column3), brand names, etc.

What I would propose instead (and how I'm planning to handle it in an unreleased tool I'm writing), is:

1. Provide a very clear definition of the syntax for extra metadata tags (the aim in this PR). This definition should fit with the minimal and whitespace-separated orientation of the todo.txt spec (eg. not add restrictions such as which characters a key could start with).

2. Implementations of tools should provide a method for users to define which keys they want interpreted as extra metadata. For example, a list of keys or a regex pattern for keys that the user are interested to track. Note that as-is, many tools only really care about specific extra metadata keys such as due: (due dates), or t: (threshold before task is shown). This would solve the time problem from #77 and the url problem, while not using the spec to restrict use cases that would be otherwise valid.

[jgclark]

Your suggestions are all making sense to me, @swalladge, with the addition of your last comment suggesting its for implementors to define actions on fields of interest. This also copes with the (English text) case of text before a colon: and after it, which technically would give a colon:<null> key:value pair, but in practice won't be acted upon. (Or at least only rarely: This is item is due: overdue, even ...)

[clach04]

@swalladge Re the empty value part you've convinced me :-) the key values really is application specific 👍 https://github.com/todotxt/todo.txt/issues/8 has some real ones seen in the wild.

When you are ready, please share your tool, I'd be interested in taking a look :-)

[samuelallan72]

Thanks @jgclark and @clach04 ! So it looks like I need to update the readme with some of the examples, and maybe the rational for how extra metadata could be implemented in tools to avoid the issues discussed above. Would you have any other concrete feedback or requested changes before you might consider merging?

When you are ready, please share your tool, I'd be interested in taking a look :-)

Absolutely! I fully intend to open source it - it's just right now it's still in early stages and I haven't had much time recently to dedicate to development. :)

[clach04]

Thanks @jgclark and @clach04 ! So it looks like I need to update the readme with some of the examples, and maybe the rational for how extra metadata could be implemented in tools to avoid the issues discussed above. Would you have any other concrete feedback or requested changes before you might consider merging?

Nothing beyond what I commented on, the PR explanation you gave was IMHO excellent and should be included in the readme as extra context.

RE merging, none of the people involved in this discussion so far have commit/merge permissions for this repo, see the open PRs (e.g. the one preceding yours). But I would merge into mine ;-)