tests for lib_string_format

[BalazsNyiro]

I collected string formatting tests to increase the cover and find differences between native Lua and LuErl

[rvirding]

Can you give me a hint in what way the string formatting differs? Is it something to do with 5.3 and integers or is it a deeper issue?

[BalazsNyiro]

The differences between LuErl and native Lua 5.3 are not too big. They are typically the number of zeros, or formatting differences.

for example in Lua:

string.format("%e, %E", math.pi, math.pi)
3.141593e+00, 3.141593E+00

In LuErl %e and %E give back the same output:
6> luerl:eval("return string.format("%e, %E", math.pi, math.pi)").
{ok,[<<"3.141593e+0, 3.141593e+0">>]}

Here are only one zero and small E in both case.

Or in other case:

string.format("%g, %g", math.pi, 10^9)
3.14159, 1e+09

Lua:

string.format("%g, %g", math.pi, 10^9)
3.14159, 1e+09

LuErl:
7> luerl:eval("return string.format("%g, %g", math.pi, 10^9)").
{ok,[<<"3.14159, 1.00000e+9">>]}

We discussed these differences with Ferenc Borocki. LuErl's results are correct if you see only the value of the results, so we think first we should collect all the differences and when we reached the 100% coverage, we can fix/debug the differences if you tell that it is important.

I think you should declare which differences should be repaired and which one's are not important.

My personal opinion is that sooner or later LuErl will be used in wider area so I think the similar behaviour would be important.

I am on holiday now but from middle of July I would be happy to reach the 100% coverage and to find the cause of all differences.

[rvirding]

OK, I will include and start attacking the problem. One reason for the difference iirc is that Luerl basically uses the builtin erlang formating and accepts what it outputs.