LittleFS as a replacement for the flash API.

[tlubke]

A couple of weeks ago I was thinking it would be nice if iii could retain all of the built-in Lua modules. In particular, the io module. If we could implement these functions with a minimal file system, it could help iii feel even more minimal.

Of course abstraction like this comes with its pros and cons. The big upsides I could foresee would be easier interfacing, and automatic wear leveling. The obvious drawback is the involved overhead cost. It looks like there have been some improvements to overhead in newer versions of LittleFS, but there is also the additional RAM cost and firmware flash space to consider.

Maybe this has all already been contemplated -- just figured I'd throw this out there while the project is in its infancy.

[tehn]

i did indeed consider this, but went hyper-minimal (block-writes to flash) in alignment with a sort of crow-aligned austerity. i did just finish up a pset_write/pset_read set of functions which write/read arbitrary tables to flash slots. there is, however, some inter-script management issues still at stake (ie, what tables are these? which script made them?)

i could see LittleFS clarifying some of these issues (ie, filenames!) but also complexifying (out of space, know what?)

performance impact is certainly a concern as i'm still getting a feel for what's possible. but i do think that using native io module would be really helpful.

[tehn]

so i checked out the link you posted and realize that overhead meant flash space.

that is the one thing that we have way too much of--- the flash is 128Mbit (16 Mbyte). for little text files that's basically infinity.

[tlubke]

Ah, I saw in #6 that the max script size was 32K and figured that a size-able portion of the remaining flash space. 16MB sounds like plenty!

RE: performance, I think the more RAM you can throw at it the better it will do on R/W speeds (up to a point). There are plenty of configuration options to play with on that front. If I had to guess, it should be comparable in most cases to what you are doing already with tables in flash slots. It could potentially be better if it is able to do some smart caching or avoid a serialization/deserialization step with tables.

I work with LittleFS on the ESP32 so I could probably help get a basic implementation running if you decide it's worth trying out.

[tehn]

thanks, that's great to know! i'm going to do some research, it's an intriguing proposal.