[tarantool-patches] Re: [PATCH 00/13] sql: support -2^63 .. 2^64-1 integer type

[n.pettik]

> On 15 Mar 2019, at 18:45, Stanislav Zudin <szudin at tarantool.org> wrote:
> 
> The patch enables support of big integers in the range [2^63, 2^64-1].
> Conversion functions use return value to inform about value
> they return - signed integer in the range [-2^63,2^63-1] or 
> unsigned integer in the range [2^63, 2^64 -1].
> The changes affect sql processing, VDBE, arithmetic functions and
> aggregate functions.

I see a lot of commits below, so it would be nice to see
some details concerning your plan of implementation.

> Issue: https://github.com/tarantool/tarantool/issues/3810
> Branch: https://github.com/tarantool/tarantool/tree/stanztt/gh-3810-sql-uint64-support

I didn’t dive into your patches yet, but I’ve come up with
simple and strange cases:

tarantool> CREATE TABLE t (id INT PRIMARY KEY, a INT);
tarantool> box.space.T:insert{1, 18446744073709551615ULL}
---
- [1, 18446744073709551615]
…
tarantool> SELECT a*1 FROM t;
---
- error: 'Failed to execute SQL statement: integer is overflowed'
...
tarantool> select a from t;
---
- - [18446744073709551615]
…

Or:

tarantool> select -9223372036854775808*(-1)
---
- error: 'Failed to execute SQL statement: integer is overflowed'
...

Second example:

tarantool> format = {}
tarantool> format[1] = {'a', 'integer'}
tarantool> format[2] = {'b', 'unsigned’}

tarantool> box.schema.space.create('T1', {format = format})
tarantool> box.space.T1:create_index('pk', {parts = {{'a'}}})

tarantool> box.sql.execute("insert into t1 values(18446744073709551615, 18446744073709551615)")
---
...

tarantool> box.sql.execute("select * from t1")
---
- - [-1, 18446744073709551615]
…

I expected that mentioned insertion would be equal to this one:

box.space.T1:insert({18446744073709551615ULL, 18446744073709551615ULL})

But instead, it is the same as:

box.space.T1:insert({18446744073709551615LL, 18446744073709551615ULL})