msgpack.NULL is not nil!

[msiomkin]

https://www.tarantool.io/en/doc/2.2/book/box/data_model/#lua-vs-msgpack

This table maps msgpack.NULL to Lua nil. It's wrong. I think there should be cdata type.

[lenkis]

@msiomkin @kyukhin let's agree on this. what exactly do we want here?

[msiomkin]

As I already said I think it should be cdata. @kyukhin right?

[lenkis]

• to add explanations as suggested in Document box.NULL #561

[Totktonada]

The table shows msgpack and lua types and also has a colon that shows examples of a Lua value. msgpack nil type has the only value: 0xc0.

Let's verify Lua -> msgpack mapping:

tarantool> msgpack.encode(nil):hex()
---
- c0
...

tarantool> msgpack.encode(box.NULL):hex()
---
- c0
...

Let's verify msgpack -> Lua mapping:

tarantool> msgpack.decode('\xc0')
---
- null
- 2
...

tarantool> type(msgpack.decode('\xc0'))
---
- cdata
...

So, yep, box.NULL (msgpack.NULL) is symmetric here, but I don't think that it would be wrong to give nil value as an example.

Anyway, Mike asks to add a type (cdata) to the column with values examples. I don't understood a motivation of this request. It at least would not be consistent: some examples would be types, some would be values.

@msiomkin Am I missed something?

[msiomkin]

some examples would be types, some would be values
@msiomkin Am I missed something?

@Totktonada "Lua type" column contains types, not values. "bin" and "array" MsgPack types map to "cdata". Why not also map msgpack.NULL to "cdata" if type(require("msgpack").NULL) equals to cdata and not nil?

[kyukhin]

I guess, @msiomkin proposes to document this:

tarantool> type(msgpack.decode('\xc0'))
---
- cdata
...

I.e. state it clearly, that box.NULL/msgpack.NULL have type cdata.