pre-compute size of serialized data

[x37v]

I'm using the slice flavor, I'm wondering if it is possible to compute the size of the serialized data without actually serializing?

[gauteh]

Hi,

I am trying out postcard as a format for storing data on a SD-card. There will be a large number of data-packages so they will exhaust the number of files allowed on FAT32. embedded-sdmmc does not (yet) support creating directories, so I can't solve it by putting the files in dirs. So now I am considering putting many packages in each file. I will use COBS between the packages, but I need to know the serialized size of the packages. The data-package is a fixed size (no variable length fields): https://github.com/gauteh/sfy/blob/main/sfy-buoy/src/axl.rs#L10 .

• Can I compute the size of a serialized package?
• If I determine it experimentally, is it stable? Will it be the the exact same size regardless of the struct contents?

Regards, Gaute

[ryankurte]

heya, i have a very similar problem and i'm not sure the macros cover the ability to pre-compute size when serializing? (or, i have missed how this can be used to achieve this).

i'm looking for something like postcard::encoded_len<T: Serialize>(t: &T) -> usize / equivalent to prost's Message::encoded_len(), that's basically a no-op encode outputting the buffer size required.

[dignifiedquire]

@jamesmunns
Any updates here? I am running into an issue with this, for preallocating buffers when serializing. The reason MAX_SIZE doesn't work is because I have &[u8] and String in my data. Maybe there could be a way to calculate the size based on MAX_SIZE + .len() dynamically?

[jamesmunns]

Reopening this, as there are actually a few different things going on here:

1. Statically computing the MAX size for a given type (for Ser/Deser)
   • This is possible already with the MAX_SIZE derive, BUT
   • This DOESN'T work for !Sized items, like slices or heap allocations. More below.
2. Dynamically computing the SPECIFIC serialized size of an object (only for Ser).
   • This I think is handled by Add function for computing the postcard serialized size of a value #66, though that has merge conflicts, which I tried to PR upstream, which never got merged: Fix Merge Conflicts parallelsystems/postcard#1
   • If you can make a new PR to postcard with the merge conflicts fixed, I will merge it

I think you are asking for item 2 above @dignifiedquire, do you want to see if that branch works for your needs, and if so you could help get that clean PR opened?

[jamesmunns]

Oops, forgot my "more below" part:

We COULD extend the MAX derive to take attributes on certain items, like:

#[derive(Serialize, Deserialize, MaxSize)]
struct Demo<'a> {
    arr: [Foo; 5],
    #[postcard(max_hint = 128)]
    sli: &'a [Bar],
    #[postcard(max_hint = 64)]
    hst: String
}

This covers the !Sized case I described above as 1b.

[dignifiedquire]

@jamesmunns just created #86 which should fix things up