polars-df 0.1.4 → 0.1.5

data/lib/polars/lazy_functions.rb

@@ -158,7 +158,7 @@ module Polars
         # TODO
         Utils.wrap_expr(_sum_exprs(exprs))
       else
-        raise Todo
+        fold(lit(0).cast(:u32), ->(a, b) { a + b }, column).alias("sum")
       end
     end
 
@@ -294,8 +294,69 @@ module Polars
       Utils.wrap_expr(RbExpr.lit(value))
     end
 
-    # def cumsum
-    # end
+    # Cumulatively sum values in a column/Series, or horizontally across list of columns/expressions.
+    #
+    # @param column [Object]
+    #   Column(s) to be used in aggregation.
+    #
+    # @return [Object]
+    #
+    # @example
+    #   df = Polars::DataFrame.new(
+    #     {
+    #       "a" => [1, 2],
+    #       "b" => [3, 4],
+    #       "c" => [5, 6]
+    #     }
+    #   )
+    #   # =>
+    #   # shape: (2, 3)
+    #   # ┌─────┬─────┬─────┐
+    #   # │ a   ┆ b   ┆ c   │
+    #   # │ --- ┆ --- ┆ --- │
+    #   # │ i64 ┆ i64 ┆ i64 │
+    #   # ╞═════╪═════╪═════╡
+    #   # │ 1   ┆ 3   ┆ 5   │
+    #   # ├╌╌╌╌╌┼╌╌╌╌╌┼╌╌╌╌╌┤
+    #   # │ 2   ┆ 4   ┆ 6   │
+    #   # └─────┴─────┴─────┘
+    #
+    # @example Cumulatively sum a column by name:
+    #   df.select(Polars.cumsum("a"))
+    #   # =>
+    #   # shape: (2, 1)
+    #   # ┌─────┐
+    #   # │ a   │
+    #   # │ --- │
+    #   # │ i64 │
+    #   # ╞═════╡
+    #   # │ 1   │
+    #   # ├╌╌╌╌╌┤
+    #   # │ 3   │
+    #   # └─────┘
+    #
+    # @example Cumulatively sum a list of columns/expressions horizontally:
+    #   df.with_column(Polars.cumsum(["a", "c"]))
+    #   # =>
+    #   # shape: (2, 4)
+    #   # ┌─────┬─────┬─────┬───────────┐
+    #   # │ a   ┆ b   ┆ c   ┆ cumsum    │
+    #   # │ --- ┆ --- ┆ --- ┆ ---       │
+    #   # │ i64 ┆ i64 ┆ i64 ┆ struct[2] │
+    #   # ╞═════╪═════╪═════╪═══════════╡
+    #   # │ 1   ┆ 3   ┆ 5   ┆ {1,6}     │
+    #   # ├╌╌╌╌╌┼╌╌╌╌╌┼╌╌╌╌╌┼╌╌╌╌╌╌╌╌╌╌╌┤
+    #   # │ 2   ┆ 4   ┆ 6   ┆ {2,8}     │
+    #   # └─────┴─────┴─────┴───────────┘
+    def cumsum(column)
+      if column.is_a?(Series)
+        column.cumsum
+      elsif column.is_a?(String)
+        col(column).cumsum
+      else
+        cumfold(lit(0).cast(:u32), ->(a, b) { a + b }, column).alias("cumsum")
+      end
+    end
 
     # Compute the spearman rank correlation between two columns.
     #
@@ -367,7 +428,7 @@ module Polars
     # def apply
     # end
 
-    # Accumulate over multiple columns horizontally/ row wise with a left fold.
+    # Accumulate over multiple columns horizontally/row wise with a left fold.
     #
     # @return [Expr]
     def fold(acc, f, exprs)
@@ -383,17 +444,118 @@ module Polars
     # def reduce
     # end
 
-    # def cumfold
-    # end
+    # Cumulatively accumulate over multiple columns horizontally/row wise with a left fold.
+    #
+    # Every cumulative result is added as a separate field in a Struct column.
+    #
+    # @param acc [Object]
+    #   Accumulator Expression. This is the value that will be initialized when the fold
+    #   starts. For a sum this could for instance be lit(0).
+    # @param f [Object]
+    #   Function to apply over the accumulator and the value.
+    #   Fn(acc, value) -> new_value
+    # @param exprs [Object]
+    #   Expressions to aggregate over. May also be a wildcard expression.
+    # @param include_init [Boolean]
+    #   Include the initial accumulator state as struct field.
+    #
+    # @return [Object]
+    #
+    # @note
+    #   If you simply want the first encountered expression as accumulator,
+    #   consider using `cumreduce`.
+    def cumfold(acc, f, exprs, include_init: false)
+      acc = Utils.expr_to_lit_or_expr(acc, str_to_lit: true)
+      if exprs.is_a?(Expr)
+        exprs = [exprs]
+      end
+
+      exprs = Utils.selection_to_rbexpr_list(exprs)
+      Utils.wrap_expr(RbExpr.cumfold(acc._rbexpr, f, exprs, include_init))
+    end
 
     # def cumreduce
     # end
 
-    # def any
-    # end
+    # Evaluate columnwise or elementwise with a bitwise OR operation.
+    #
+    # @return [Expr]
+    def any(name)
+      if name.is_a?(String)
+        col(name).any
+      else
+        fold(lit(false), ->(a, b) { a.cast(:bool) | b.cast(:bool) }, name).alias("any")
+      end
+    end
 
-    # def exclude
-    # end
+    # Exclude certain columns from a wildcard/regex selection.
+    #
+    # @param columns [Object]
+    #   Column(s) to exclude from selection
+    #   This can be:
+    #
+    #   - a column name, or multiple column names
+    #   - a regular expression starting with `^` and ending with `$`
+    #   - a dtype or multiple dtypes
+    #
+    # @return [Object]
+    #
+    # @example
+    #   df = Polars::DataFrame.new(
+    #     {
+    #       "aa" => [1, 2, 3],
+    #       "ba" => ["a", "b", nil],
+    #       "cc" => [nil, 2.5, 1.5]
+    #     }
+    #   )
+    #   # =>
+    #   # shape: (3, 3)
+    #   # ┌─────┬──────┬──────┐
+    #   # │ aa  ┆ ba   ┆ cc   │
+    #   # │ --- ┆ ---  ┆ ---  │
+    #   # │ i64 ┆ str  ┆ f64  │
+    #   # ╞═════╪══════╪══════╡
+    #   # │ 1   ┆ a    ┆ null │
+    #   # ├╌╌╌╌╌┼╌╌╌╌╌╌┼╌╌╌╌╌╌┤
+    #   # │ 2   ┆ b    ┆ 2.5  │
+    #   # ├╌╌╌╌╌┼╌╌╌╌╌╌┼╌╌╌╌╌╌┤
+    #   # │ 3   ┆ null ┆ 1.5  │
+    #   # └─────┴──────┴──────┘
+    #
+    # @example Exclude by column name(s):
+    #   df.select(Polars.exclude("ba"))
+    #   # =>
+    #   # shape: (3, 2)
+    #   # ┌─────┬──────┐
+    #   # │ aa  ┆ cc   │
+    #   # │ --- ┆ ---  │
+    #   # │ i64 ┆ f64  │
+    #   # ╞═════╪══════╡
+    #   # │ 1   ┆ null │
+    #   # ├╌╌╌╌╌┼╌╌╌╌╌╌┤
+    #   # │ 2   ┆ 2.5  │
+    #   # ├╌╌╌╌╌┼╌╌╌╌╌╌┤
+    #   # │ 3   ┆ 1.5  │
+    #   # └─────┴──────┘
+    #
+    # @example Exclude by regex, e.g. removing all columns whose names end with the letter "a":
+    #   df.select(Polars.exclude("^.*a$"))
+    #   # =>
+    #   # shape: (3, 1)
+    #   # ┌──────┐
+    #   # │ cc   │
+    #   # │ ---  │
+    #   # │ f64  │
+    #   # ╞══════╡
+    #   # │ null │
+    #   # ├╌╌╌╌╌╌┤
+    #   # │ 2.5  │
+    #   # ├╌╌╌╌╌╌┤
+    #   # │ 1.5  │
+    #   # └──────┘
+    def exclude(columns)
+      col("*").exclude(columns)
+    end
 
     # Do one of two things.
     #
@@ -429,11 +591,26 @@ module Polars
       end
     end
 
-    # def groups
-    # end
+    # Syntactic sugar for `Polars.col("foo").agg_groups`.
+    #
+    # @return [Object]
+    def groups(column)
+      col(column).agg_groups
+    end
 
-    # def quantile
-    # end
+    # Syntactic sugar for `Polars.col("foo").quantile(...)`.
+    #
+    # @param column [String]
+    #   Column name.
+    # @param quantile [Float]
+    #   Quantile between 0.0 and 1.0.
+    # @param interpolation ["nearest", "higher", "lower", "midpoint", "linear"]
+    #   Interpolation method.
+    #
+    # @return [Expr]
+    def quantile(column, quantile, interpolation: "nearest")
+      col(column).quantile(quantile, interpolation: interpolation)
+    end
 
     # Create a range expression (or Series).
     #
@@ -449,7 +626,7 @@ module Polars
     # @param eager [Boolean]
     #   If eager evaluation is `True`, a Series is returned instead of an Expr.
     # @param dtype [Symbol]
-    #   Apply an explicit integer dtype to the resulting expression (default is Int64).
+    #   Apply an explicit integer dtype to the resulting expression (default is `:i64`).
     #
     # @return [Expr, Series]
     #
@@ -474,14 +651,212 @@ module Polars
       end
     end
 
-    # def argsort_by
-    # end
+    # Find the indexes that would sort the columns.
+    #
+    # Argsort by multiple columns. The first column will be used for the ordering.
+    # If there are duplicates in the first column, the second column will be used to
+    # determine the ordering and so on.
+    #
+    # @param exprs [Object]
+    #   Columns use to determine the ordering.
+    # @param reverse [Boolean]
+    #   Default is ascending.
+    #
+    # @return [Expr]
+    def argsort_by(exprs, reverse: false)
+      if !exprs.is_a?(Array)
+        exprs = [exprs]
+      end
+      if reverse == true || reverse == false
+        reverse = [reverse] * exprs.length
+      end
+      exprs = Utils.selection_to_rbexpr_list(exprs)
+      Utils.wrap_expr(RbExpr.argsort_by(exprs, reverse))
+    end
 
-    # def duration
-    # end
+    # Create polars `Duration` from distinct time components.
+    #
+    # @return [Expr]
+    #
+    # @example
+    #   df = Polars::DataFrame.new(
+    #     {
+    #       "datetime" => [DateTime.new(2022, 1, 1), DateTime.new(2022, 1, 2)],
+    #       "add" => [1, 2]
+    #     }
+    #   )
+    #   df.select(
+    #     [
+    #       (Polars.col("datetime") + Polars.duration(weeks: "add")).alias("add_weeks"),
+    #       (Polars.col("datetime") + Polars.duration(days: "add")).alias("add_days"),
+    #       (Polars.col("datetime") + Polars.duration(seconds: "add")).alias("add_seconds"),
+    #       (Polars.col("datetime") + Polars.duration(milliseconds: "add")).alias(
+    #         "add_milliseconds"
+    #       ),
+    #       (Polars.col("datetime") + Polars.duration(hours: "add")).alias("add_hours")
+    #     ]
+    #   )
+    #   # =>
+    #   # shape: (2, 5)
+    #   # ┌─────────────────────┬─────────────────────┬─────────────────────┬─────────────────────────┬─────────────────────┐
+    #   # │ add_weeks           ┆ add_days            ┆ add_seconds         ┆ add_milliseconds        ┆ add_hours           │
+    #   # │ ---                 ┆ ---                 ┆ ---                 ┆ ---                     ┆ ---                 │
+    #   # │ datetime[ns]        ┆ datetime[ns]        ┆ datetime[ns]        ┆ datetime[ns]            ┆ datetime[ns]        │
+    #   # ╞═════════════════════╪═════════════════════╪═════════════════════╪═════════════════════════╪═════════════════════╡
+    #   # │ 2022-01-08 00:00:00 ┆ 2022-01-02 00:00:00 ┆ 2022-01-01 00:00:01 ┆ 2022-01-01 00:00:00.001 ┆ 2022-01-01 01:00:00 │
+    #   # ├╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌┼╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌┼╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌┼╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌┼╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌┤
+    #   # │ 2022-01-16 00:00:00 ┆ 2022-01-04 00:00:00 ┆ 2022-01-02 00:00:02 ┆ 2022-01-02 00:00:00.002 ┆ 2022-01-02 02:00:00 │
+    #   # └─────────────────────┴─────────────────────┴─────────────────────┴─────────────────────────┴─────────────────────┘
+    def duration(
+      days: nil,
+      seconds: nil,
+      nanoseconds: nil,
+      microseconds: nil,
+      milliseconds: nil,
+      minutes: nil,
+      hours: nil,
+      weeks: nil
+    )
+      if !hours.nil?
+        hours = Utils.expr_to_lit_or_expr(hours, str_to_lit: false)._rbexpr
+      end
+      if !minutes.nil?
+        minutes = Utils.expr_to_lit_or_expr(minutes, str_to_lit: false)._rbexpr
+      end
+      if !seconds.nil?
+        seconds = Utils.expr_to_lit_or_expr(seconds, str_to_lit: false)._rbexpr
+      end
+      if !milliseconds.nil?
+        milliseconds = Utils.expr_to_lit_or_expr(milliseconds, str_to_lit: false)._rbexpr
+      end
+      if !microseconds.nil?
+        microseconds = Utils.expr_to_lit_or_expr(microseconds, str_to_lit: false)._rbexpr
+      end
+      if !nanoseconds.nil?
+        nanoseconds = Utils.expr_to_lit_or_expr(nanoseconds, str_to_lit: false)._rbexpr
+      end
+      if !days.nil?
+        days = Utils.expr_to_lit_or_expr(days, str_to_lit: false)._rbexpr
+      end
+      if !weeks.nil?
+        weeks = Utils.expr_to_lit_or_expr(weeks, str_to_lit: false)._rbexpr
+      end
 
-    # def format
-    # end
+      Utils.wrap_expr(
+        _rb_duration(
+          days,
+          seconds,
+          nanoseconds,
+          microseconds,
+          milliseconds,
+          minutes,
+          hours,
+          weeks
+        )
+      )
+    end
+
+    # Horizontally concat Utf8 Series in linear time. Non-Utf8 columns are cast to Utf8.
+    #
+    # @param exprs [Object]
+    #   Columns to concat into a Utf8 Series.
+    # @param sep [String]
+    #   String value that will be used to separate the values.
+    #
+    # @return [Expr]
+    #
+    # @example
+    #   df = Polars::DataFrame.new(
+    #     {
+    #       "a" => [1, 2, 3],
+    #       "b" => ["dogs", "cats", nil],
+    #       "c" => ["play", "swim", "walk"]
+    #     }
+    #   )
+    #   df.with_columns(
+    #     [
+    #       Polars.concat_str(
+    #         [
+    #           Polars.col("a") * 2,
+    #           Polars.col("b"),
+    #           Polars.col("c")
+    #         ],
+    #         sep: " "
+    #       ).alias("full_sentence")
+    #     ]
+    #   )
+    #   # =>
+    #   # shape: (3, 4)
+    #   # ┌─────┬──────┬──────┬───────────────┐
+    #   # │ a   ┆ b    ┆ c    ┆ full_sentence │
+    #   # │ --- ┆ ---  ┆ ---  ┆ ---           │
+    #   # │ i64 ┆ str  ┆ str  ┆ str           │
+    #   # ╞═════╪══════╪══════╪═══════════════╡
+    #   # │ 1   ┆ dogs ┆ play ┆ 2 dogs play   │
+    #   # ├╌╌╌╌╌┼╌╌╌╌╌╌┼╌╌╌╌╌╌┼╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌┤
+    #   # │ 2   ┆ cats ┆ swim ┆ 4 cats swim   │
+    #   # ├╌╌╌╌╌┼╌╌╌╌╌╌┼╌╌╌╌╌╌┼╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌┤
+    #   # │ 3   ┆ null ┆ walk ┆ null          │
+    #   # └─────┴──────┴──────┴───────────────┘
+    def concat_str(exprs, sep: "")
+      exprs = Utils.selection_to_rbexpr_list(exprs)
+      return Utils.wrap_expr(RbExpr.concat_str(exprs, sep))
+    end
+
+    # Format expressions as a string.
+    #
+    # @param fstring [String]
+    #   A string that with placeholders.
+    #   For example: "hello_{}" or "{}_world
+    # @param args [Object]
+    #   Expression(s) that fill the placeholders
+    #
+    # @return [Expr]
+    #
+    # @example
+    #   df = Polars::DataFrame.new(
+    #     {
+    #       "a": ["a", "b", "c"],
+    #       "b": [1, 2, 3]
+    #     }
+    #   )
+    #   df.select(
+    #     [
+    #       Polars.format("foo_{}_bar_{}", Polars.col("a"), "b").alias("fmt")
+    #     ]
+    #   )
+    #   # =>
+    #   # shape: (3, 1)
+    #   # ┌─────────────┐
+    #   # │ fmt         │
+    #   # │ ---         │
+    #   # │ str         │
+    #   # ╞═════════════╡
+    #   # │ foo_a_bar_1 │
+    #   # ├╌╌╌╌╌╌╌╌╌╌╌╌╌┤
+    #   # │ foo_b_bar_2 │
+    #   # ├╌╌╌╌╌╌╌╌╌╌╌╌╌┤
+    #   # │ foo_c_bar_3 │
+    #   # └─────────────┘
+    def format(fstring, *args)
+      if fstring.scan("{}").length != args.length
+        raise ArgumentError, "number of placeholders should equal the number of arguments"
+      end
+
+      exprs = []
+
+      arguments = args.each
+      fstring.split(/(\{\})/).each do |s|
+        if s == "{}"
+          e = Utils.expr_to_lit_or_expr(arguments.next, str_to_lit: false)
+          exprs << e
+        elsif s.length > 0
+          exprs << lit(s)
+        end
+      end
+
+      concat_str(exprs, sep: "")
+    end
 
     # Concat the arrays in a Series dtype List in linear time.
     #
@@ -491,8 +866,73 @@ module Polars
       Utils.wrap_expr(RbExpr.concat_lst(exprs))
     end
 
-    # def collect_all
-    # end
+    # Collect multiple LazyFrames at the same time.
+    #
+    # This runs all the computation graphs in parallel on Polars threadpool.
+    #
+    # @param lazy_frames [Boolean]
+    #   A list of LazyFrames to collect.
+    # @param type_coercion [Boolean]
+    #   Do type coercion optimization.
+    # @param predicate_pushdown [Boolean]
+    #   Do predicate pushdown optimization.
+    # @param projection_pushdown [Boolean]
+    #   Do projection pushdown optimization.
+    # @param simplify_expression [Boolean]
+    #   Run simplify expressions optimization.
+    # @param string_cache [Boolean]
+    #   This argument is deprecated and will be ignored
+    # @param no_optimization [Boolean]
+    #   Turn off optimizations.
+    # @param slice_pushdown [Boolean]
+    #   Slice pushdown optimization.
+    # @param common_subplan_elimination [Boolean]
+    #   Will try to cache branching subplans that occur on self-joins or unions.
+    # @param allow_streaming [Boolean]
+    #   Run parts of the query in a streaming fashion (this is in an alpha state)
+    #
+    # @return [Array]
+    def collect_all(
+      lazy_frames,
+      type_coercion: true,
+      predicate_pushdown: true,
+      projection_pushdown: true,
+      simplify_expression: true,
+      string_cache: false,
+      no_optimization: false,
+      slice_pushdown: true,
+      common_subplan_elimination: true,
+      allow_streaming: false
+    )
+      if no_optimization
+        predicate_pushdown = false
+        projection_pushdown = false
+        slice_pushdown = false
+        common_subplan_elimination = false
+      end
+
+      prepared = []
+
+      lazy_frames.each do |lf|
+        ldf = lf._ldf.optimization_toggle(
+          type_coercion,
+          predicate_pushdown,
+          projection_pushdown,
+          simplify_expression,
+          slice_pushdown,
+          common_subplan_elimination,
+          allow_streaming
+        )
+        prepared << ldf
+      end
+
+      out = _collect_all(prepared)
+
+      # wrap the rbdataframes into dataframe
+      result = out.map { |rbdf| Utils.wrap_df(rbdf) }
+
+      result
+    end
 
     # Run polars expressions without a context.
     #
@@ -621,11 +1061,105 @@ module Polars
       end
     end
 
-    # def coalesce
-    # end
+    # Folds the expressions from left to right, keeping the first non-null value.
+    #
+    # @param exprs [Object]
+    #   Expressions to coalesce.
+    #
+    # @return [Expr]
+    #
+    # @example
+    #   df = Polars::DataFrame.new(
+    #     [
+    #       [nil, 1.0, 1.0],
+    #       [nil, 2.0, 2.0],
+    #       [nil, nil, 3.0],
+    #       [nil, nil, nil]
+    #     ],
+    #     columns: [["a", :f64], ["b", :f64], ["c", :f64]]
+    #   )
+    #   df.with_column(Polars.coalesce(["a", "b", "c", 99.9]).alias("d"))
+    #   # =>
+    #   # shape: (4, 4)
+    #   # ┌──────┬──────┬──────┬──────┐
+    #   # │ a    ┆ b    ┆ c    ┆ d    │
+    #   # │ ---  ┆ ---  ┆ ---  ┆ ---  │
+    #   # │ f64  ┆ f64  ┆ f64  ┆ f64  │
+    #   # ╞══════╪══════╪══════╪══════╡
+    #   # │ null ┆ 1.0  ┆ 1.0  ┆ 1.0  │
+    #   # ├╌╌╌╌╌╌┼╌╌╌╌╌╌┼╌╌╌╌╌╌┼╌╌╌╌╌╌┤
+    #   # │ null ┆ 2.0  ┆ 2.0  ┆ 2.0  │
+    #   # ├╌╌╌╌╌╌┼╌╌╌╌╌╌┼╌╌╌╌╌╌┼╌╌╌╌╌╌┤
+    #   # │ null ┆ null ┆ 3.0  ┆ 3.0  │
+    #   # ├╌╌╌╌╌╌┼╌╌╌╌╌╌┼╌╌╌╌╌╌┼╌╌╌╌╌╌┤
+    #   # │ null ┆ null ┆ null ┆ 99.9 │
+    #   # └──────┴──────┴──────┴──────┘
+    def coalesce(exprs)
+      exprs = Utils.selection_to_rbexpr_list(exprs)
+      Utils.wrap_expr(_coalesce_exprs(exprs))
+    end
 
-    # def from_epoch
-    # end
+    # Utility function that parses an epoch timestamp (or Unix time) to Polars Date(time).
+    #
+    # Depending on the `unit` provided, this function will return a different dtype:
+    # - unit: "d" returns pl.Date
+    # - unit: "s" returns pl.Datetime["us"] (pl.Datetime's default)
+    # - unit: "ms" returns pl.Datetime["ms"]
+    # - unit: "us" returns pl.Datetime["us"]
+    # - unit: "ns" returns pl.Datetime["ns"]
+    #
+    # @param column [Object]
+    #     Series or expression to parse integers to pl.Datetime.
+    # @param unit [String]
+    #     The unit of the timesteps since epoch time.
+    # @param eager [Boolean]
+    #     If eager evaluation is `true`, a Series is returned instead of an Expr.
+    #
+    # @return [Object]
+    #
+    # @example
+    #   df = Polars::DataFrame.new({"timestamp" => [1666683077, 1666683099]}).lazy
+    #   df.select(Polars.from_epoch(Polars.col("timestamp"), unit: "s")).collect
+    #   # =>
+    #   # shape: (2, 1)
+    #   # ┌─────────────────────┐
+    #   # │ timestamp           │
+    #   # │ ---                 │
+    #   # │ datetime[μs]        │
+    #   # ╞═════════════════════╡
+    #   # │ 2022-10-25 07:31:17 │
+    #   # ├╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌╌┤
+    #   # │ 2022-10-25 07:31:39 │
+    #   # └─────────────────────┘
+    def from_epoch(column, unit: "s", eager: false)
+      if column.is_a?(String)
+        column = col(column)
+      elsif !column.is_a?(Series) && !column.is_a?(Expr)
+        column = Series.new(column)
+      end
+
+      if unit == "d"
+        expr = column.cast(:date)
+      elsif unit == "s"
+        raise Todo
+        # expr = (column.cast(:i64) * 1_000_000).cast(Datetime("us"))
+      elsif Utils::DTYPE_TEMPORAL_UNITS.include?(unit)
+        raise Todo
+        # expr = column.cast(Datetime(unit))
+      else
+        raise ArgumentError, "'unit' must be one of {{'ns', 'us', 'ms', 's', 'd'}}, got '#{unit}'."
+      end
+
+      if eager
+        if !column.is_a?(Series)
+          raise ArgumentError, "expected Series or Array if eager: true, got #{column.class.name}"
+        else
+          column.to_frame.select(expr).to_series
+        end
+      else
+        expr
+      end
+    end
 
     # Start a "when, then, otherwise" expression.
     #

data/lib/polars/rolling_group_by.rb

@@ -0,0 +1,35 @@
+module Polars
+  # A rolling grouper.
+  #
+  # This has an `.agg` method which will allow you to run all polars expressions in a
+  # groupby context.
+  class RollingGroupBy
+    def initialize(
+      df,
+      index_column,
+      period,
+      offset,
+      closed,
+      by
+    )
+      period = Utils._timedelta_to_pl_duration(period)
+      offset = Utils._timedelta_to_pl_duration(offset)
+
+      @df = df
+      @time_column = index_column
+      @period = period
+      @offset = offset
+      @closed = closed
+      @by = by
+    end
+
+    def agg(aggs)
+      @df.lazy
+        .groupby_rolling(
+          index_column: @time_column, period: @period, offset: @offset, closed: @closed, by: @by
+        )
+        .agg(aggs)
+        .collect(no_optimization: true, string_cache: false)
+    end
+  end
+end