Price comparison weirdness

[jmjatlanta]

I am not sure if this is a "feature" or a bug...

I am here talking about the price object and it's operator <

I assume that the price object compares a ratio, and not the actual numbers. But the code indicates that it first compares the instance IDs, and then compares the product of base and quote of the two objects.

Anomaly 1: I would assume that given asset object with the IDs of 1 and 2, the price does not play into it at all. 1 is less than 2, regardless of the price. So to do a legit comparison, you must assure the base and quote of the two are the same. Otherwise you will get an ID comparison and not price.

Anomaly 2: The prices are compared by multiplication instead of ratios. Hence:

      const auto& a = create_bitasset("ASSETA");
      const auto& b = create_bitasset("ASSETB");

      // now compare IDs
      price p1( asset( 100, a.id), asset( 100, b.id) );
      price p2( asset( 100, b.id), asset( 100, a.id) );

      BOOST_CHECK( p1 < p2 ); // checks ids, and sees P1's base ID is less than P2's
      BOOST_CHECK( p2 > p1 ); // the opposite should also be true

      // now compare prices
      price p3( asset( 1, a.id), asset( 2, b.id) );
      price p4( asset( 100, a.id), asset( 200, b.id) );

      // those prices are equal, but it doesn't come out that way
      BOOST_CHECK( p1 < p2 );
      BOOST_CHECK( p2 > p1 );

passes. Even though prices P3 and P4 are in essence equal. Is this a bug, or simply must be kept in mind as an implementation detail?

Perhaps a related tangent: #1094

[abitmore]

These lines of code appeared twice in your test case.

      BOOST_CHECK( p1 < p2 );
      BOOST_CHECK( p2 > p1 );

Prices are compared by asset IDs first because we need to index them with by_price tag in databases, to get orders on the same side of a market one by one. So the direction (base vs quote) matters.

[jmjatlanta]

Stupid mistake. Thanks.