Are there any disadvantages to changing id format to match extant files?

[mediapathic]

Hey all,

First off, many thanks to Prot and everyone else for your work on this project. It has solved so many thorny problems for me already, and I am deeply grateful.

To my question:

I have a long history of extant note files (in the multiple thousands), mostly markdown and some org. Many of them are already named (or internally identified in metadata -- I've gone through a lot of systems) with a UID that looks like YYYYMMDDhhmm, e.g. 202506021557 (hi @DivineDominion this is your fault :). I am considering trying to change the ids generated by denote to match this. I currently have only a few dozen notes created with the default ids, and it seems like I should decide on a format before I have too many that I will need to change, possibly manually, in the future.

My question is, are there any significant long-term downsides to do this? The only one I see immediately is losing the granularity of seconds, but that is a problem I have run into maybe three times in the past many years, and I don't mind fudging by a minute to resolve that when it does happen.

I'm also considering if it's worth the effort to bother. If I make any old notes identifiable to denote it'll require renaming the files anyway, as far too many of them have spaces in the filename etc. My main concern there is that by changing e.g. 202506021605 into 20250602T160500 I'm going to lose the searchability of that reference in other systems outside of denote (I use these IDs everywhere). But as I write that out I realize that's not a question anyone but me can answer.

Still, though, my question remains: are there any reasons you can see to not change the IDs that denote produces to the format I'm already using? And are the advantages worth the trouble?

Thank you!

[protesilaos]

Hello @mediapathic!

Still, though, my question remains: are there any reasons you can see to not change the IDs that denote produces to the format I'm already using? And are the advantages worth the trouble?

To me, the Denote identifier is a decent compromise between descriptiveness and uniqueness. It will fail if you need to create unique file names at the nanosecond level, but will cope perfectly with the more common scenario of writing a file every few seconds, minutes, hours, or days.

If I make any old notes identifiable to denote it'll require renaming the files anyway, as far too many of them have spaces in the filename etc.

The spaces should not be a problem. The only essential component is the identifier.

My main concern there is that by changing e.g. 202506021605 into 20250602T160500 I'm going to lose the searchability of that reference in other systems outside of denote (I use these IDs everywhere). But as I write that out I realize that's not a question anyone but me can answer.

Depending on how you go about performing the search, this can be achieved with some regular expression. Here is an Emacs-style regexp that matches both: "20250602T?1605\\([0-9][0-9]\\)?". You can test it by writing the identifiers to a buffer and then using M-x re-builder.

At any rate, @jeanphilippegg has outlined the challenges of making custom identifiers a user option. See issue 581 and specifically this comment: #581 (comment). To quote:

It is not simple to allow the customization of the identifier format. We have to consider the following:

• How to handle uniqueness of ids, particularly when the custom format may not include the seconds.
• Ids are used for links. They should not be allowed to include square brackets or parenthesis, or else links will break.
• Custom ids must be fontified properly in Dired buffers.
• We must be able to parse the file name, no matter the configuration at any point in time. For example, given "@@1234--my-note.org", we must always be able to deduce that "1234" is the identifier, even if the user decide to come back to the classic identifier format.

As you can see, it is not simple, but I believe it is doable. I have a few ideas and I will look at this.

[jeanphilippegg]

My question is, are there any significant long-term downsides to do this? The only one I see immediately is losing the granularity of seconds, but that is a problem I have run into maybe three times in the past many years, and I don't mind fudging by a minute to resolve that when it does happen.

Still, though, my question remains: are there any reasons you can see to not change the IDs that denote produces to the format I'm already using? And are the advantages worth the trouble?

I am working on it. I should make a pull request in that direction soon. At that point, I would not see a disadvantage in using your custom format. But it is currently not possible.

By the way, in my config, I have this: (setq denote-id-format "%Y%m%dT%H%M00"). This makes the second always 00, because I also find them unnecessary. Denote automatically makes ids unique to the minute. However, denote-id-format is a defcustom and so it is not intended to be modified. But I know it works.