CB-05: String Operations

In this part of the cookbook, we are going to use the cleaned dataset from the previous part:

(def weather-2012
  (g/read-csv! "data/cookbook/weather-2012.csv"))

5.1 Finding The Snowiest Months

Spark (and Geni by extension) has many string operations. To filter for rows that contains "snow" in the weather column in a case insensitive way, we use g/lower to make the string all lower case and g/like to find the "snow" pattern:

(-> weather-2012
    (g/filter (g/like (g/lower :weather) "%snow%"))
    (g/select :weather)
    g/distinct
    g/show)
; +--------------------------+
; |weather                   |
; +--------------------------+
; |Rain,Snow                 |
; |Freezing Rain,Snow Grains |
; |Snow,Fog                  |
; |Rain Showers,Snow Showers |
; |Moderate Snow,Blowing Snow|
; |Snow Showers              |
; |Drizzle,Snow,Fog          |
; |Moderate Snow             |
; |Snow,Haze                 |
; |Rain,Snow,Fog             |
; |Snow Pellets              |
; |Rain,Snow,Ice Pellets     |
; |Snow Showers,Fog          |
; |Snow,Blowing Snow         |
; |Rain,Snow Grains          |
; |Snow,Ice Pellets          |
; |Snow                      |
; |Freezing Drizzle,Snow     |
; |Drizzle,Snow              |
; +--------------------------+

We verify that every distinct row contains the word "snow". To find the snowiest month, we do the usual group-by + aggregate operation to count the distinct dates of every weather:

(-> weather-2012
    (g/filter (g/like (g/lower :weather) "%snow%"))
    (g/group-by :year :month)
    (g/agg {:n-days (g/count-distinct :day)})
    (g/order-by :year :month)
    g/show)
; +----+-----+------+
; |year|month|n-days|
; +----+-----+------+
; |2012|1    |22    |
; |2012|2    |19    |
; |2012|3    |10    |
; |2012|4    |3     |
; |2012|11   |6     |
; |2012|12   |22    |
; +----+-----+------+

We find that January and December are indeed the snowiest months in Montréal.

5.2 Putting Snowiness and Temperature Together

To determine the snowiness of a month, we compute the proportion of snowy days to the total number of days in that month. To count the number of snowy days, we count the number of distinct days only for snowy days. To achieve this we use g/when with only the then-clause, namely (g/when (g/like (g/lower :weather) "%snow%") :day). This says that if the row's weather contains "snow", then it takes the value of the column :day, otherwise it is null.

(-> weather-2012
    (g/group-by :year :month)
    (g/agg
      {:n-snow-days (g/count-distinct
                     (g/when (g/like (g/lower :weather) "%snow%") :day))
       :n-days      (g/count-distinct :day)
       :mean-temp   (g/mean :temp)})
    (g/order-by :year :month)
    (g/select {:year      :year
               :month     :month
               :snowiness (g/format-number (g// :n-snow-days :n-days) 2)
               :mean-temp (g/format-number :mean-temp 1)})
    g/show)
; +----+-----+---------+---------+
; |year|month|snowiness|mean-temp|
; +----+-----+---------+---------+
; |2012|1    |0.71     |-7.4     |
; |2012|2    |0.66     |-4.2     |
; |2012|3    |0.32     |3.1      |
; |2012|4    |0.10     |7.0      |
; |2012|5    |0.00     |16.2     |
; |2012|6    |0.00     |20.1     |
; |2012|7    |0.00     |22.8     |
; |2012|8    |0.00     |22.3     |
; |2012|9    |0.00     |16.5     |
; |2012|10   |0.00     |11.0     |
; |2012|11   |0.20     |0.9      |
; |2012|12   |0.71     |-3.3     |
; +----+-----+---------+---------+

Notice that we used g/format-number to specify the number of decimal places to show.

5.3 Finding Temperatures of Common Weather Descriptions

In each row, the :weather value is a comma-separated weather description. To separate the individual weather descriptions, we can use (g/split :weather ",") to get a column of list of strings. We further invoke g/explode on this column, which is analogous to the flat-map operation or Clojure's mapcat:

(-> weather-2012
    (g/with-column :weather-description (g/explode (g/split :weather ",")))
    (g/group-by :weather-description)
    (g/agg {:mean-temp (g/mean :temp)
            :n-days    (g/count-distinct :year :month :day)})
    (g/order-by (g/desc :mean-temp))
    (g/select {:weather-description :weather-description
               :mean-temp (g/format-number :mean-temp 1)
               :n-days :n-days})
    (g/show {:num-rows 25}))
; +---------------------+---------+------+
; |weather-description  |mean-temp|n-days|
; +---------------------+---------+------+
; |Thunderstorms        |20.2     |14    |
; |Moderate Rain Showers|19.6     |1     |
; |Rain Showers         |14.2     |99    |
; |Mainly Clear         |12.6     |288   |
; |Heavy Rain Showers   |10.9     |1     |
; |Mostly Cloudy        |10.6     |315   |
; |Rain                 |9.0      |86    |
; |Cloudy               |8.0      |285   |
; |Drizzle              |6.9      |38    |
; |Clear                |6.8      |216   |
; |Fog                  |5.0      |77    |
; |Moderate Rain        |1.7      |1     |
; |Snow Pellets         |0.7      |1     |
; |Ice Pellets          |-0.7     |6     |
; |Haze                 |-1.2     |7     |
; |Snow Grains          |-1.6     |2     |
; |Freezing Rain        |-3.7     |7     |
; |Snow Showers         |-3.8     |25    |
; |Snow                 |-4.1     |71    |
; |Freezing Drizzle     |-4.7     |12    |
; |Blowing Snow         |-5.4     |5     |
; |Moderate Snow        |-5.5     |2     |
; |Freezing Fog         |-7.6     |3     |
; +---------------------+---------+------+