higher level dsl

[sckott]

As @richfitz mentioned, rex provides a good example of how we could build a dsl to define jq queries. @richfitz you like the lazyeval approach? want to use here?

[richfitz]

dplyr would be the other place to look, as many of the verbs are going to be the same. Unfortunately I don't really use the package so can't really comment. lazyeval seems like the right call, probably using https://github.com/hadley/lazyeval

[sckott]

@richfitz started hacking on dsl a bit, e.g.,

library("httr")
library("jsonlite")
out <- minify(content(GET('https://api.github.com/repos/stedolan/jq/commits?per_page=5'), "text"))

# the jq call, and equivalent with dsl
# jq '.'
out %>% dot() %>% jq
# jq '.[]'
out %>% index() %>% jq
# jq '.[0]'
out %>% index(0) %>% jq
# jq '.[1]'
out %>% index(1) %>% jq
# jq '.[0] | {message: .commit.message, name: .commit.committer.name}'
out %>% index(0) %>% select(message = .commit.message, name = .commit.committer.name) %>% jq
# jq '.[] | {message: .commit.message, name: .commit.committer.name}'
out %>% index() %>% select(message = .commit.message, name = .commit.committer.name) %>% jq

[richfitz]

Are there ways of testing if the operator is the last in a chain? Then we could omit the jq() call e.g.? But this whole chaining thing is pretty new to me :smiley:

[sckott]

good question @richfitz - i have asked Stefan before, - maybe that SO thread has an answer - I'll try that, would be nice to get rid of jq() call at the end.

With rex, it seems that they don't pipe operations together, but instead compose one big query specification. Do you think that's a better approach? I think probably not, but that would be easier to develop i guess

[sckott]

this is fully implemented now, i'll open a separate issue asking for feedback on the DSL