tests: Make all table driven tests tell you where they failed

[ValarDragon]

Currently the majority of our table driven tests don't tell you where exactly they failed. (This is true of most tests in types, random example: https://github.com/cosmos/cosmos-sdk/blob/master/types/rational_test.go#L57) This makes debugging extremely hard. We definitely need to do something in the spirit of telling you which test case faiiled, however this doesn't need to be prelaunch, could easily be pre1.0.

Here is an example of a test case in something I created:

require.Equal(t, expectedTx[:32], updatedTx[:32],
       "First 32 bytes of the txs differed. testcase #%d, tx test #%d", tcIndex, i))

This makes it easy to figure out what you're looking at (as opposed to just a pile of different bytes in a file with many tests), and exactly which test case failed.

The above is actually not ideal. Here I propose is a more ideal way which we standardize throughout the sdk.

require.Equal(t, obj1, obj2, 
  "<desc>. tc #%d, i #%d", test case index, other variables necessary to specify at what step this occured

(Newline doesn't matter, was just including for ease of reading in the PR)

where <desc> is a short description which indicates what the objs are (since you don't have variable names), and what is failing. The indices that specify where it occured can be as simple as "i" and "j" in the for loop, or the variable name.

A sample log output of the above would be:

--- FAIL: TestGenerateTxUpdateTxConsistentency (0.53s)
	require.go:157: 
			Error Trace:	transacter_test.go:44
			Error:      	Not equal: 
			            	expected: []byte{0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xcb, 0x6, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xde, 0xec, 0xab, 0x16, 0x49, 0xd, 0x38, 0xc0, 0x91, 0x97, 0x6b, 0x93, 0x1c, 0x41, 0x6d, 0xd, 0xa5, 0xe9, 0xa0, 0xda, 0x5, 0x0, 0x0, 0x0}
			            	actual  : []byte{0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xcb, 0x6, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0xde, 0xec, 0xab, 0x16, 0x49, 0xd, 0x38, 0xc0, 0x91, 0x97, 0x6b, 0x93, 0x1c, 0x41, 0x6d, 0xd, 0xa4, 0xe9, 0xa0, 0xda, 0x5, 0x0, 0x0, 0x0}
			            	
			            	Diff:
			            	--- Expected
			            	+++ Actual
			            	@@ -3,3 +3,3 @@
			            	  00000010  de ec ab 16 49 0d 38 c0  91 97 6b 93 1c 41 6d 0d  |....I.8...k..Am.|
			            	- 00000020  a5 e9 a0 da 05 00 00 00                           |........|
			            	+ 00000020  a4 e9 a0 da 05 00 00 00                           |........|
			            	 }
			Test:       	TestGenerateTxUpdateTxConsistentency
			Messages:   	First 32 bytes of the txs differed. tc #2, i #793

The idea is that it should be easy to go from failed test case to where exactly the error occured, as line number doesn't tell you what test case, or what part of the test case. (In many cases, you run a for loop some amount of iterations) This would save debugging time significantly, when you're breaking an already existing thing.

[rigelrozanski]

Big ++ to this - super frustrating

[alexanderbez]

Good idea, let's start doing this!

[ValarDragon]

Lets document this somewhere in contributing.md, and require this of all new tests that get written in future PR's.

[rigelrozanski]

@ValarDragon want to add to contributing.md and then close this issue?

[ValarDragon]

Sure!