What does it mean if Z statistic is -Inf? I have ran a Wilcoxon rank sum test to compare two independent samples. I now want to get a Z statistic to report, but when I do so, it results in -Inf. What does this mean in regards to reporting results? ...code below.

I have mydata and want to compare scores in two classes, A and B.
> head(mydata)

         score class
18001 1.019240     A
18002 1.020216     A
18003 1.021493     A
18005 1.022327     A
18006 1.018774     A
18007 1.021493     A

> results <- wilcox.test(score ~ class, data = mydata, na.rm = TRUE)
> results

    Wilcoxon rank sum test with continuity correction

data:  score by class
W = 9129975, p-value < 2.2e-16
alternative hypothesis: true location shift is not equal to 0

> qnorm(results$p.value) #convert p-value to Z statistic
[1] -Inf

 A: It means what you should have known before, namely that the p-value of the Wilcoxon test is essentially zero (apparently zero up to machine precision), so small that the function that you use to compute the corresponding normal quantile gives you -Inf. The null hypothesis of the test is rejected as strong as a rejection can be, corresponding with an infinite value of what you call "Z-statistic" (your Wilcoxon test is two-sided and you Z-statistic computation makes no allowance for that, which seems wrong, however the Inf wouldn't go away anyway).
In the first place I don't think it's a good idea to convert the p-value into a Z-statistic. The p-value has a very clear interpretation so why would you want to have this converted into a Z-statistic, which you then have problems to make sense of? 
A: The wilcox.test function actually computes the z value, but then doesn't include it in the output.  If you know some R coding, it's pretty easy to modify the function's code to print out the z value.  I made a convenience function in the rcompanion package to do this.
It looks like this approach is better than using qnorm in that it returns the correct sign for z, and outputs a finite answer for very small p values.
require(rcompanion)

C = 30000:40000
D = 10000:20000

wilcoxonZ(C, D)

   ###    z 
   ###  122 

wilcox.test(C, D)

   ### W = 100020001, p-value < 2.2e-16

qnorm(wilcox.test(C, D)$p.value)

   ### [1] -Inf

Another approach is to use the output from the coin package, which will return similar results.
require(coin)

Value = c(C, D)
Group = c(rep("C", length(C)), rep("D", length(D))) 
Data  = data.frame(Group, Value)

wilcox_test(Value ~ Group, data=Data)

   ### Asymptotic Wilcoxon-Mann-Whitney Test
   ### 
   ### Z = 122.48, p-value < 2.2e-16

