Use HAVING for summarise() + filter()

[mgirlich]

This solves the first of two cases where HAVING can be used according to #870.

library(dplyr, warn.conflicts = FALSE)
library(dbplyr, warn.conflicts = FALSE)

lazy_frame(g = 1, x = 1, y = 1) %>% 
  group_by(g) %>% 
  summarise(x_mean = mean(x, na.rm = TRUE)) %>% 
  filter(x_mean > 1, g == 1)

Before

SELECT *
FROM (
  SELECT `g`, AVG(`x`) AS `x_mean`
  FROM `df`
  GROUP BY `g`
) `q01`
WHERE (`x_mean` > 1.0) AND (`g` = 1.0)

After

SELECT `g`, AVG(`x`) AS `x_mean`
FROM `df`
GROUP BY `g`
HAVING (AVG(`x`) > 1.0) AND (`g` = 1.0)

[hadley]

This is looking great!