Good morning 🙋‍♂️ - Is there a technical reason for the inability to sort by a text field? I imagine dynamically doing it is much slower but I thought the index would handle that at indexing time. Or perhaps I'm crazy and you can do alphabetical sorts ?

Yes, we've discussed this here: https://github.com/typesense/typesense/issues/117

Would it be possible, technically speaking, to sort at indexing time and create a sort of "virtual" numerical field on the item which is used as a proxy to sort alphabetically ?

One way to do this with existing feature set will be to convert your string to a lexographically ordered integer. For e.g. given a string "apple", you can create an int64 value that just concatenates the 8-but ascii representation of the individual letters: a, p, p, l, e

had already written this so I'm sending it 🙂 but mostly for confirmation as I think you answered my question I feel like I'm missing something in my understanding. In our case we have a regular "name" field which is maybe 30 characters at most. We could call it a "title" field to better illustrate the type of field. So before actually sending the documents to typesense for indexing, is there anything preventing me from pre-sorting in my code and creating a numerical value from 1 to X based on the output of that sort ?

Yes, you can just pre-sort and index that way. In fact you can even pre-sort multiple fields as long as you have 1 corresponding numerical value that represents the document's position in that particular sort order.

Ah there's the downside! 😄 Hadn't thought of that but it's true. Any update to the title field requires all items to be reindexed. In our case we push items to the database and have a listener/subscriber on

@AfterInsert

and

@AfterUpdate

events which keeps Typesense in sync. We'd need to add a trigger which updates the sort on all items if the title field is modified on one of them. It works better for our case because despite having millions of documents, they are segregated cleanly into about 5-10k chunks. As for supporting sorting on any string length, it doesn't seem worth it. Algolia certainly discourage people from doing it except in rare circumstances and even then as a separate index or something. Sorting by a fixed length title sort of field could be interesting but I'm not sure how to determine what an optimal length is. 8 characters is probably good enough because at a glance I doubt most people can mentally sort past a certain number of characters, although in titles starting with the same word it would be easier to identify something that's not sorted correctly past the first 8 characters.

Yes, that's what I think too. Technically English ASCII is just 7 bits, so with 64 bits you get 9 chars 🙂 It is mostly useful for eye-balling data as you said.

Good point! Makes it quite a bit more difficult in those scenarios. Either way, I believe you have quite a few things on your roadmap with a higher priority than this. Fun problem to solve though 👌

We are looking forward to string sorting. (Need to sort the names of German courts alphabetically)