For nested joins, there was an issue where calls to the inner most so…

[quinnj]

…urce got translated like :(a.(b.c)) while proper access should look like :((a.b).c). This commit adds an invert_getproperty_access translate function that performs the translation.

Fixes a query like:

using Query, DataFrames
mf = (senID=[1,2,3], leftNeighbor=[1,2,3], linearPosition=[1,2,3], word=["as", "the", "sailor"], simplecat=[1,2,3])
@from i in mf begin
    @join i11 in mf on (i.senID, i.leftNeighbor) equals (i11.senID, i11.linearPosition)
    @join i12 in mf on (i11.senID, i11.leftNeighbor) equals (i12.senID, i12.linearPosition)
    @join i13 in mf on (i12.senID, i12.leftNeighbor) equals (i13.senID, i13.linearPosition)
    @where i13.word == "as"
    @group i by i.simplecat into g
    @select {Key = key(g), Count = length(g)}
    @collect DataFrame
end

[codecov-io]

Codecov Report

Merging #211 into master will decrease coverage by 1.02%.
The diff coverage is 66.66%.

@@            Coverage Diff            @@
##           master    #211      +/-   ##
=========================================
- Coverage   92.22%   91.2%   -1.03%     
=========================================
  Files           5       5              
  Lines         296     307      +11     
=========================================
+ Hits          273     280       +7     
- Misses         23      27       +4

Impacted Files	Coverage Δ
src/query_translation.jl	91.45% <66.66%> (-1.14%)	⬇️

---

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 7900bbc...6222079. Read the comment docs.

[quinnj]

@davidanthoff, this fixes the original queries Irene shared on slack (both queries with 3 joins & 5 joins). It seems to work on a handful of others I tried as well. The existing test suite also passes, so I don't think it breaks anything else. I added a few simple unit tests for the shift_access! function. It sounded on slack that there might be other issues this also helps? Maybe some of those queries can all be pooled together to add some more tests. Feel free to push/update this branch if you'd like.

[davidanthoff]

This is fantastic, it also fixes #133! And many bug reports I got via other channels that I never even filed here because I always thought I had to fix the inference stuff to sort this out :)

I added one test with multiple @let statements that I think is the simplest case that gets at the core of this issue.

You mentioned some unit tests for shift_access!, but I think you might not have pushed them? Do you want to add those, and then I'll merge?

[quinnj]

Ah, thanks for the reminder. I thought I had pushed the tests, but they were just sitting here locally. I just pushed them.

[davidanthoff]

Cool. I'll wait for CI to pass and then I'll merge.