Port TPL behavior for float literals. Change print behavior for REAL.

[lmwnshn]

Heading

Port TPL behavior for float literals. Change print behavior for REAL.

Description

@tpan496 found a bug where values outside of float range were treated as 0 or inf.

We were casting doubles down into floats in the bytecode_generator.
Something we missed when porting TPL2 and making the "64-bit instead of 32-bit by default" change.

It's another one of those "fixed in TPL" things so I went and yoinked that code.

This also changes the default behavior of printing real values, e.g., SELECT 0.3333;.

• A. The old behavior of to_string is to always print to 6 decimal places.
  • Failure case: prints 0.333333333 as 0.333333.
• B. The fmt lib will print as many decimal places as are available, but will not be able to print trailing 0s.
  • Failure case: prints 0.333000 as 0.333.
• C. If we want to print trailing 0s, we can use stringstream and std::fixed.
  • Failure case: prints 0.333 as 0.333000.

In general, there is no good solution here. Given a REAL, we don't know if it was float or double initially. If precision matters, revive the decimal PR.
Going with B. for now.

I hope this does not blow up float SQL-based tests...

[noisepage-checks]

Major Decrease in Performance

STOP: this PR has a major negative performance impact

tps (%change)	benchmark_type	wal_device	details
-2.61%	tpcc	RAM disk	Details master tps=22554.9, commit tps=21967.0, query_mode=extended, benchmark_type=tpcc, scale_factor=32.0000, terminals=32, client_time=60, weights={'Payment': 43, 'Delivery': 4, 'NewOrder': 45, 'StockLevel': 4, 'OrderStatus': 4}, wal_device=RAM disk, max_connection_threads=32
1.13%	tpcc	None	Details master tps=28904.05, commit tps=29229.8, query_mode=extended, benchmark_type=tpcc, scale_factor=32.0000, terminals=32, client_time=60, weights={'Payment': 43, 'Delivery': 4, 'NewOrder': 45, 'StockLevel': 4, 'OrderStatus': 4}, wal_device=None, max_connection_threads=32
-2.18%	tpcc	HDD	Details master tps=21953.18, commit tps=21474.99, query_mode=extended, benchmark_type=tpcc, scale_factor=32.0000, terminals=32, client_time=60, weights={'Payment': 43, 'Delivery': 4, 'NewOrder': 45, 'StockLevel': 4, 'OrderStatus': 4}, wal_device=HDD, max_connection_threads=32
-2.23%	tatp	RAM disk	Details master tps=6500.06, commit tps=6355.35, query_mode=extended, benchmark_type=tatp, scale_factor=1.0000, terminals=16, client_time=60, weights={'GetAccessData': 35, 'UpdateLocation': 14, 'GetNewDestination': 10, 'GetSubscriberData': 35, 'DeleteCallForwarding': 2, 'InsertCallForwarding': 2, 'UpdateSubscriberData': 2}, wal_device=RAM disk, max_connection_threads=32
-7.35%	tatp	None	Details master tps=7421.06, commit tps=6875.95, query_mode=extended, benchmark_type=tatp, scale_factor=1.0000, terminals=16, client_time=60, weights={'GetAccessData': 35, 'UpdateLocation': 14, 'GetNewDestination': 10, 'GetSubscriberData': 35, 'DeleteCallForwarding': 2, 'InsertCallForwarding': 2, 'UpdateSubscriberData': 2}, wal_device=None, max_connection_threads=32
1.49%	tatp	HDD	Details master tps=6446.34, commit tps=6542.3, query_mode=extended, benchmark_type=tatp, scale_factor=1.0000, terminals=16, client_time=60, weights={'GetAccessData': 35, 'UpdateLocation': 14, 'GetNewDestination': 10, 'GetSubscriberData': 35, 'DeleteCallForwarding': 2, 'InsertCallForwarding': 2, 'UpdateSubscriberData': 2}, wal_device=HDD, max_connection_threads=32

[codecov]

Codecov Report

Merging #1586 (b3d8fc1) into master (7f83ef8) will increase coverage by 0.41%.
The diff coverage is 80.00%.

@@            Coverage Diff             @@
##           master    #1586      +/-   ##
==========================================
+ Coverage   81.32%   81.73%   +0.41%     
==========================================
  Files         706      739      +33     
  Lines       49981    52045    +2064     
==========================================
+ Hits        40645    42541    +1896     
- Misses       9336     9504     +168     

Impacted Files	Coverage Δ
src/execution/vm/bytecode_generator.cpp	84.20% <75.00%> (-0.03%)	⬇️
src/network/postgres/postgres_packet_writer.cpp	75.87% <100.00%> (ø)
src/network/network_io_wrapper.cpp	82.25% <0.00%> (-3.23%)	⬇️
src/include/storage/sql_table.h	97.67% <0.00%> (-2.33%)	⬇️
src/execution/sema/sema_decl.cpp	79.24% <0.00%> (-1.89%)	⬇️
src/storage/data_table.cpp	97.15% <0.00%> (-1.14%)	⬇️
src/include/catalog/index_schema.h	95.96% <0.00%> (-0.79%)	⬇️
src/execution/sema/sema_expr.cpp	76.13% <0.00%> (-0.57%)	⬇️
src/execution/sql/sorter.cpp	96.73% <0.00%> (-0.55%)	⬇️
src/execution/compiler/codegen.cpp	87.48% <0.00%> (-0.24%)	⬇️
... and 50 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 7f83ef8...b3d8fc1. Read the comment docs.

[tpan496]

LGTM. Should we be worried about the performance guard?

[lmwnshn]

Nah, that thing is super noisy.