hardened handling of aliases #73
Conversation
jameslamb
force-pushed
the
bugfix/get_fields_aliases
branch
from
62dca64
to
1b8691e
Compare
|
fun fact ES5.x onwards and legacy ES versions return alias results slightly differently. Added a fix to this PR. I'm glad we caught it before the release |
jameslamb
force-pushed
the
bugfix/get_fields_aliases
branch
from
1b8691e
to
9989e04
Compare
Codecov Report
@@ Coverage Diff @@
## master #73 +/- ##
==========================================
+ Coverage 86.08% 86.93% +0.84%
==========================================
Files 9 9
Lines 611 643 +32
==========================================
+ Hits 526 559 +33
+ Misses 85 84 -1
Continue to review full report at Codecov.
|
| , mappingDT = mappingDT | ||
| ) | ||
| , fill = TRUE | ||
| ) |
There was a problem hiding this comment.
I think there's a better way you could do this with a data.table merge but these DTs are small anyway so it probably doesn't matter
There was a problem hiding this comment.
yeah they're pretty small and this makes it really explicit. This is an EDA function so super-super fast performance is less important than correctness, IMHO
| return(.process_alias(alias_string = resultContent)) | ||
| major_version <- .get_es_version(es_host) | ||
| process_alias <- switch( | ||
| major_version |
There was a problem hiding this comment.
since there are also minor changes in scrolling between the versions i'd love if we could pull all the version-specific stuff up higher and configure w/ passing functions instead of putting internal switches like this
doesn't have to be a blocker for now though since thats a pretty large and maybe unnecessary change for our purposes
There was a problem hiding this comment.
fair point! I will throw that up in an issue
Sorry @austin3dickey , one more that I want to go out on the v0.3.0 release.
We don't have any tests right now on how aliases are handled in
get_fields. This PR addresses that.One thing to call out is that having multiple aliases point to the same index is a totally valid thing, and our code previously couldn't handle that. This PR adds that support as well.
One other minor bugfix...
urlis a base function, so I changed our internal references toes_url. I noticed it because I was getting the dreadedclosure not subsettableerror while debugging.