Fix bug checking the length of binary types

[p32blo]

When we have a restriction on a binary type to check for its length, the size that needs to be considered is the octet size.

So for a xs:hexBinary that might be represented by a string of 4 caracteres it actually only has length 2.

The same for xs:base64Binary but the computation of the length is not as straight forward.

Here is the relevant part of the standard:

lenght - https://www.w3.org/TR/xmlschema11-2/#rf-length

For hexBinary and base64Binary and datatypes ·derived· from them, length is measured in octets (8 bits) of binary data.

hexBinary - https://www.w3.org/TR/xmlschema11-2/#hexBinary

The ·value space· of hexBinary is the set of finite-length sequences of zero or more binary octets.

base64Binary - https://www.w3.org/TR/xmlschema11-2/#base64Binary

The length of a base64Binary value may be calculated from the ·lexical representation· by removing whitespace and padding characters and performing the calculation shown in the pseudo-code below:

lex2   := killwhitespace(lexform)    -- remove whitespace characters
lex3   := strip_equals(lex2)         -- strip padding characters at end
length := floor (length(lex3) * 3 / 4)         -- calculate length

This is my attempt to fix this, but feel free to change it completely if it does not make sense to do it this way :)

[Bergmann89]

Hey @p32blo,

thanks for your contribution. The fix looks good in principle, but I’d
suggest a slightly different approach. Handling special constraints
directly in the generator shifts complexity to the wrong place. In my
opinion, it’s better to let the type itself handle this.

I would introduce the following types (in xsd-parser-types::xml):

pub struct HexString(pub String);

impl HexString {
    /// Returns the byte length of the decoded data, and not the length of the string.
    pub fn len(&self) -> usize { ... }
    pub fn as_str(&self) -> &str { ... }
}

pub struct Base64String(pub String);

impl Base64String {
    /// Returns the byte length of the decoded data, and not the length of the string.
    pub fn len(&self) -> usize { ... }
    pub fn as_str(&self) -> &str { ... }
}

pub struct HexBinary(pub Vec<u8>);

impl Deref for HexBinary { ... }

pub struct Base64Binary(pub Vec<u8>);

impl Deref for Base64Binary { ... }

By default, HexString and Base64String would represent
xs:hexBinary and xs:base64Binary. This keeps compatibility with the
current string-based approach and avoids decoding unless necessary. If
a user needs the decoded form, they can switch to the binary types.

Providing suitable helpers in the config (see
Config::with_qname_type and Config::with_qname_type_from) would let
users choose the type they want.

This approach should not require changes to the interpreter or
generator, keeping the codebase cleaner while still solving the general
problem.

[p32blo]

Thanks for the feedback. I tried to give it a go :)
Is this closer to what you envision?

I was not able to figure out how to make the HexString and Base64String the default and I don't know how the existing Base64Binary in xsd-parser-types::xml should intersect with this new one, if at all. I will leave that to your consideration.

[Bergmann89]

Hey @p32blo,

Sorry for the late response. Yes, this looks way more cleaner as the first version 👍

I have added some small remarks, to finalize it.

You can find the default definition here. You can just replace it with the newly defined types like shown here.

[p32blo]

I think I addressed all the comments.

I'm still not sure if it's the best way to handle the default values of xs:hexBinary and xs:base64Binary so feel free to change it !