polars-df 0.1.4 → 0.2.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/CHANGELOG.md +13 -0
- data/Cargo.lock +430 -217
- data/Cargo.toml +2 -0
- data/LICENSE.txt +1 -1
- data/README.md +0 -2
- data/ext/polars/Cargo.toml +9 -3
- data/ext/polars/src/apply/dataframe.rs +303 -0
- data/ext/polars/src/apply/mod.rs +253 -0
- data/ext/polars/src/apply/series.rs +1173 -0
- data/ext/polars/src/conversion.rs +254 -35
- data/ext/polars/src/dataframe.rs +151 -6
- data/ext/polars/src/error.rs +8 -0
- data/ext/polars/src/lazy/apply.rs +34 -2
- data/ext/polars/src/lazy/dataframe.rs +80 -3
- data/ext/polars/src/lazy/dsl.rs +84 -10
- data/ext/polars/src/lib.rs +180 -8
- data/ext/polars/src/series.rs +328 -10
- data/ext/polars/src/utils.rs +25 -0
- data/lib/polars/convert.rb +100 -0
- data/lib/polars/data_frame.rb +1480 -77
- data/lib/polars/data_types.rb +122 -0
- data/lib/polars/date_time_expr.rb +10 -10
- data/lib/polars/date_time_name_space.rb +8 -8
- data/lib/polars/dynamic_group_by.rb +52 -0
- data/lib/polars/expr.rb +262 -12
- data/lib/polars/functions.rb +194 -5
- data/lib/polars/group_by.rb +76 -36
- data/lib/polars/io.rb +19 -3
- data/lib/polars/lazy_frame.rb +798 -25
- data/lib/polars/lazy_functions.rb +569 -30
- data/lib/polars/list_expr.rb +1 -1
- data/lib/polars/rolling_group_by.rb +35 -0
- data/lib/polars/series.rb +192 -27
- data/lib/polars/string_expr.rb +6 -5
- data/lib/polars/string_name_space.rb +1 -1
- data/lib/polars/utils.rb +25 -8
- data/lib/polars/version.rb +1 -1
- data/lib/polars.rb +38 -29
- metadata +11 -4
data/Cargo.toml
CHANGED
@@ -3,6 +3,8 @@ members = ["ext/polars"]
|
|
3
3
|
|
4
4
|
[patch.crates-io]
|
5
5
|
jsonpath_lib = { git = "https://github.com/ritchie46/jsonpath", rev = "24eaf0b4416edff38a4d1b6b17bc4b9f3f047b4b" }
|
6
|
+
halfbrown = { git = "https://github.com/Licenser/halfbrown", rev = "952023c5dd6461b009bb5ba66b9aa979bd75949f" }
|
7
|
+
arrow2 = { git = "https://github.com/ankane/arrow2", rev = "9f36b2b97446e6dd495473e4361a70d863ac8027" }
|
6
8
|
|
7
9
|
[profile.release]
|
8
10
|
strip = true
|
data/LICENSE.txt
CHANGED
data/README.md
CHANGED
@@ -12,8 +12,6 @@ Add this line to your application’s Gemfile:
|
|
12
12
|
gem "polars-df"
|
13
13
|
```
|
14
14
|
|
15
|
-
Note: Rust is currently required for installation, and it can take 15-20 minutes to compile the extension.
|
16
|
-
|
17
15
|
## Getting Started
|
18
16
|
|
19
17
|
This library follows the [Polars Python API](https://pola-rs.github.io/polars/py-polars/html/reference/index.html).
|
data/ext/polars/Cargo.toml
CHANGED
@@ -1,6 +1,7 @@
|
|
1
1
|
[package]
|
2
2
|
name = "polars"
|
3
|
-
version = "0.
|
3
|
+
version = "0.2.0"
|
4
|
+
license = "MIT"
|
4
5
|
authors = ["Andrew Kane <andrew@ankane.org>"]
|
5
6
|
edition = "2021"
|
6
7
|
publish = false
|
@@ -9,16 +10,19 @@ publish = false
|
|
9
10
|
crate-type = ["cdylib"]
|
10
11
|
|
11
12
|
[dependencies]
|
13
|
+
ahash = "0.8"
|
12
14
|
magnus = "0.4"
|
13
|
-
polars-core = "0.
|
15
|
+
polars-core = "0.26.1"
|
14
16
|
serde_json = "1"
|
15
17
|
|
16
18
|
[dependencies.polars]
|
17
|
-
version = "0.
|
19
|
+
version = "0.26.1"
|
18
20
|
features = [
|
19
21
|
"abs",
|
20
22
|
"arange",
|
21
23
|
"arg_where",
|
24
|
+
"asof_join",
|
25
|
+
"avro",
|
22
26
|
"concat_str",
|
23
27
|
"cse",
|
24
28
|
"csv-file",
|
@@ -53,6 +57,7 @@ features = [
|
|
53
57
|
"partition_by",
|
54
58
|
"pct_change",
|
55
59
|
"performant",
|
60
|
+
"pivot",
|
56
61
|
"product",
|
57
62
|
"propagate_nans",
|
58
63
|
"random",
|
@@ -61,6 +66,7 @@ features = [
|
|
61
66
|
"repeat_by",
|
62
67
|
"rolling_window",
|
63
68
|
"round_series",
|
69
|
+
"row_hash",
|
64
70
|
"search_sorted",
|
65
71
|
"semi_anti_join",
|
66
72
|
"serde-lazy",
|
@@ -0,0 +1,303 @@
|
|
1
|
+
use magnus::{class, RArray, TryConvert, Value};
|
2
|
+
use polars::prelude::*;
|
3
|
+
use polars_core::frame::row::{rows_to_schema_first_non_null, Row};
|
4
|
+
use polars_core::series::SeriesIter;
|
5
|
+
|
6
|
+
use super::*;
|
7
|
+
use crate::{RbDataFrame, RbPolarsErr, RbSeries, Wrap};
|
8
|
+
|
9
|
+
fn get_iters(df: &DataFrame) -> Vec<SeriesIter> {
|
10
|
+
df.get_columns().iter().map(|s| s.iter()).collect()
|
11
|
+
}
|
12
|
+
|
13
|
+
fn get_iters_skip(df: &DataFrame, skip: usize) -> Vec<std::iter::Skip<SeriesIter>> {
|
14
|
+
df.get_columns()
|
15
|
+
.iter()
|
16
|
+
.map(|s| s.iter().skip(skip))
|
17
|
+
.collect()
|
18
|
+
}
|
19
|
+
|
20
|
+
pub fn apply_lambda_unknown<'a>(
|
21
|
+
df: &'a DataFrame,
|
22
|
+
lambda: Value,
|
23
|
+
inference_size: usize,
|
24
|
+
) -> RbResult<(Value, bool)> {
|
25
|
+
let mut null_count = 0;
|
26
|
+
let mut iters = get_iters(df);
|
27
|
+
|
28
|
+
for _ in 0..df.height() {
|
29
|
+
let iter = iters.iter_mut().map(|it| Wrap(it.next().unwrap()));
|
30
|
+
let arg = (iter.collect::<Vec<Wrap<AnyValue>>>(),);
|
31
|
+
let out: Value = lambda.funcall("call", arg)?;
|
32
|
+
|
33
|
+
if out.is_nil() {
|
34
|
+
null_count += 1;
|
35
|
+
continue;
|
36
|
+
} else if out.is_kind_of(class::true_class()) || out.is_kind_of(class::false_class()) {
|
37
|
+
let first_value = out.try_convert::<bool>().ok();
|
38
|
+
return Ok((
|
39
|
+
RbSeries::new(
|
40
|
+
apply_lambda_with_bool_out_type(df, lambda, null_count, first_value)
|
41
|
+
.into_series(),
|
42
|
+
)
|
43
|
+
.into(),
|
44
|
+
false,
|
45
|
+
));
|
46
|
+
} else if out.is_kind_of(class::float()) {
|
47
|
+
let first_value = out.try_convert::<f64>().ok();
|
48
|
+
|
49
|
+
return Ok((
|
50
|
+
RbSeries::new(
|
51
|
+
apply_lambda_with_primitive_out_type::<Float64Type>(
|
52
|
+
df,
|
53
|
+
lambda,
|
54
|
+
null_count,
|
55
|
+
first_value,
|
56
|
+
)
|
57
|
+
.into_series(),
|
58
|
+
)
|
59
|
+
.into(),
|
60
|
+
false,
|
61
|
+
));
|
62
|
+
} else if out.is_kind_of(class::integer()) {
|
63
|
+
let first_value = out.try_convert::<i64>().ok();
|
64
|
+
return Ok((
|
65
|
+
RbSeries::new(
|
66
|
+
apply_lambda_with_primitive_out_type::<Int64Type>(
|
67
|
+
df,
|
68
|
+
lambda,
|
69
|
+
null_count,
|
70
|
+
first_value,
|
71
|
+
)
|
72
|
+
.into_series(),
|
73
|
+
)
|
74
|
+
.into(),
|
75
|
+
false,
|
76
|
+
));
|
77
|
+
// } else if out.is_kind_of(class::string()) {
|
78
|
+
// let first_value = out.try_convert::<String>().ok();
|
79
|
+
// return Ok((
|
80
|
+
// RbSeries::new(
|
81
|
+
// apply_lambda_with_utf8_out_type(df, lambda, null_count, first_value)
|
82
|
+
// .into_series(),
|
83
|
+
// )
|
84
|
+
// .into(),
|
85
|
+
// false,
|
86
|
+
// ));
|
87
|
+
} else if out.respond_to("_s", true)? {
|
88
|
+
let rb_rbseries: Value = out.funcall("_s", ()).unwrap();
|
89
|
+
let series = rb_rbseries
|
90
|
+
.try_convert::<&RbSeries>()
|
91
|
+
.unwrap()
|
92
|
+
.series
|
93
|
+
.borrow();
|
94
|
+
let dt = series.dtype();
|
95
|
+
return Ok((
|
96
|
+
RbSeries::new(
|
97
|
+
apply_lambda_with_list_out_type(df, lambda, null_count, Some(&series), dt)?
|
98
|
+
.into_series(),
|
99
|
+
)
|
100
|
+
.into(),
|
101
|
+
false,
|
102
|
+
));
|
103
|
+
} else if out.try_convert::<Wrap<Row<'a>>>().is_ok() {
|
104
|
+
let first_value = out.try_convert::<Wrap<Row<'a>>>().unwrap().0;
|
105
|
+
return Ok((
|
106
|
+
RbDataFrame::from(
|
107
|
+
apply_lambda_with_rows_output(
|
108
|
+
df,
|
109
|
+
lambda,
|
110
|
+
null_count,
|
111
|
+
first_value,
|
112
|
+
inference_size,
|
113
|
+
)
|
114
|
+
.map_err(RbPolarsErr::from)?,
|
115
|
+
)
|
116
|
+
.into(),
|
117
|
+
true,
|
118
|
+
));
|
119
|
+
} else if out.is_kind_of(class::array()) {
|
120
|
+
return Err(RbPolarsErr::other(
|
121
|
+
"A list output type is invalid. Do you mean to create polars List Series?\
|
122
|
+
Then return a Series object."
|
123
|
+
.into(),
|
124
|
+
));
|
125
|
+
} else {
|
126
|
+
return Err(RbPolarsErr::other("Could not determine output type".into()));
|
127
|
+
}
|
128
|
+
}
|
129
|
+
Err(RbPolarsErr::other("Could not determine output type".into()))
|
130
|
+
}
|
131
|
+
|
132
|
+
fn apply_iter<T>(
|
133
|
+
df: &DataFrame,
|
134
|
+
lambda: Value,
|
135
|
+
init_null_count: usize,
|
136
|
+
skip: usize,
|
137
|
+
) -> impl Iterator<Item = Option<T>> + '_
|
138
|
+
where
|
139
|
+
T: TryConvert,
|
140
|
+
{
|
141
|
+
let mut iters = get_iters_skip(df, init_null_count + skip);
|
142
|
+
((init_null_count + skip)..df.height()).map(move |_| {
|
143
|
+
let iter = iters.iter_mut().map(|it| Wrap(it.next().unwrap()));
|
144
|
+
let tpl = (iter.collect::<Vec<Wrap<AnyValue>>>(),);
|
145
|
+
match lambda.funcall::<_, _, Value>("call", tpl) {
|
146
|
+
Ok(val) => val.try_convert::<T>().ok(),
|
147
|
+
Err(e) => panic!("ruby function failed {}", e),
|
148
|
+
}
|
149
|
+
})
|
150
|
+
}
|
151
|
+
|
152
|
+
/// Apply a lambda with a primitive output type
|
153
|
+
pub fn apply_lambda_with_primitive_out_type<D>(
|
154
|
+
df: &DataFrame,
|
155
|
+
lambda: Value,
|
156
|
+
init_null_count: usize,
|
157
|
+
first_value: Option<D::Native>,
|
158
|
+
) -> ChunkedArray<D>
|
159
|
+
where
|
160
|
+
D: RbArrowPrimitiveType,
|
161
|
+
D::Native: Into<Value> + TryConvert,
|
162
|
+
{
|
163
|
+
let skip = usize::from(first_value.is_some());
|
164
|
+
if init_null_count == df.height() {
|
165
|
+
ChunkedArray::full_null("apply", df.height())
|
166
|
+
} else {
|
167
|
+
let iter = apply_iter(df, lambda, init_null_count, skip);
|
168
|
+
iterator_to_primitive(iter, init_null_count, first_value, "apply", df.height())
|
169
|
+
}
|
170
|
+
}
|
171
|
+
|
172
|
+
/// Apply a lambda with a boolean output type
|
173
|
+
pub fn apply_lambda_with_bool_out_type(
|
174
|
+
df: &DataFrame,
|
175
|
+
lambda: Value,
|
176
|
+
init_null_count: usize,
|
177
|
+
first_value: Option<bool>,
|
178
|
+
) -> ChunkedArray<BooleanType> {
|
179
|
+
let skip = usize::from(first_value.is_some());
|
180
|
+
if init_null_count == df.height() {
|
181
|
+
ChunkedArray::full_null("apply", df.height())
|
182
|
+
} else {
|
183
|
+
let iter = apply_iter(df, lambda, init_null_count, skip);
|
184
|
+
iterator_to_bool(iter, init_null_count, first_value, "apply", df.height())
|
185
|
+
}
|
186
|
+
}
|
187
|
+
|
188
|
+
/// Apply a lambda with utf8 output type
|
189
|
+
pub fn apply_lambda_with_utf8_out_type(
|
190
|
+
df: &DataFrame,
|
191
|
+
lambda: Value,
|
192
|
+
init_null_count: usize,
|
193
|
+
first_value: Option<&str>,
|
194
|
+
) -> Utf8Chunked {
|
195
|
+
let skip = usize::from(first_value.is_some());
|
196
|
+
if init_null_count == df.height() {
|
197
|
+
ChunkedArray::full_null("apply", df.height())
|
198
|
+
} else {
|
199
|
+
let iter = apply_iter::<String>(df, lambda, init_null_count, skip);
|
200
|
+
iterator_to_utf8(iter, init_null_count, first_value, "apply", df.height())
|
201
|
+
}
|
202
|
+
}
|
203
|
+
|
204
|
+
/// Apply a lambda with list output type
|
205
|
+
pub fn apply_lambda_with_list_out_type<'a>(
|
206
|
+
df: &'a DataFrame,
|
207
|
+
lambda: Value,
|
208
|
+
init_null_count: usize,
|
209
|
+
first_value: Option<&Series>,
|
210
|
+
dt: &DataType,
|
211
|
+
) -> RbResult<ListChunked> {
|
212
|
+
let skip = usize::from(first_value.is_some());
|
213
|
+
if init_null_count == df.height() {
|
214
|
+
Ok(ChunkedArray::full_null("apply", df.height()))
|
215
|
+
} else {
|
216
|
+
let mut iters = get_iters_skip(df, init_null_count + skip);
|
217
|
+
let iter = ((init_null_count + skip)..df.height()).map(|_| {
|
218
|
+
let iter = iters.iter_mut().map(|it| Wrap(it.next().unwrap()));
|
219
|
+
let tpl = (iter.collect::<Vec<Wrap<AnyValue>>>(),);
|
220
|
+
match lambda.funcall::<_, _, Value>("call", tpl) {
|
221
|
+
Ok(val) => match val.funcall::<_, _, Value>("_s", ()) {
|
222
|
+
Ok(val) => val
|
223
|
+
.try_convert::<&RbSeries>()
|
224
|
+
.ok()
|
225
|
+
.map(|ps| ps.series.borrow().clone()),
|
226
|
+
Err(_) => {
|
227
|
+
if val.is_nil() {
|
228
|
+
None
|
229
|
+
} else {
|
230
|
+
panic!("should return a Series, got a {:?}", val)
|
231
|
+
}
|
232
|
+
}
|
233
|
+
},
|
234
|
+
Err(e) => panic!("ruby function failed {}", e),
|
235
|
+
}
|
236
|
+
});
|
237
|
+
iterator_to_list(dt, iter, init_null_count, first_value, "apply", df.height())
|
238
|
+
}
|
239
|
+
}
|
240
|
+
|
241
|
+
pub fn apply_lambda_with_rows_output<'a>(
|
242
|
+
df: &'a DataFrame,
|
243
|
+
lambda: Value,
|
244
|
+
init_null_count: usize,
|
245
|
+
first_value: Row<'a>,
|
246
|
+
inference_size: usize,
|
247
|
+
) -> PolarsResult<DataFrame> {
|
248
|
+
let width = first_value.0.len();
|
249
|
+
let null_row = Row::new(vec![AnyValue::Null; width]);
|
250
|
+
|
251
|
+
let mut row_buf = Row::default();
|
252
|
+
|
253
|
+
let skip = 1;
|
254
|
+
let mut iters = get_iters_skip(df, init_null_count + skip);
|
255
|
+
let mut row_iter = ((init_null_count + skip)..df.height()).map(|_| {
|
256
|
+
let iter = iters.iter_mut().map(|it| Wrap(it.next().unwrap()));
|
257
|
+
let tpl = (iter.collect::<Vec<Wrap<AnyValue>>>(),);
|
258
|
+
match lambda.funcall::<_, _, Value>("call", tpl) {
|
259
|
+
Ok(val) => {
|
260
|
+
match val.try_convert::<RArray>().ok() {
|
261
|
+
Some(tuple) => {
|
262
|
+
row_buf.0.clear();
|
263
|
+
for v in tuple.each() {
|
264
|
+
let v = v.unwrap().try_convert::<Wrap<AnyValue>>().unwrap().0;
|
265
|
+
row_buf.0.push(v);
|
266
|
+
}
|
267
|
+
let ptr = &row_buf as *const Row;
|
268
|
+
// Safety:
|
269
|
+
// we know that row constructor of polars dataframe does not keep a reference
|
270
|
+
// to the row. Before we mutate the row buf again, the reference is dropped.
|
271
|
+
// we only cannot prove it to the compiler.
|
272
|
+
// we still do this because it saves a Vec allocation in a hot loop.
|
273
|
+
unsafe { &*ptr }
|
274
|
+
}
|
275
|
+
None => &null_row,
|
276
|
+
}
|
277
|
+
}
|
278
|
+
Err(e) => panic!("ruby function failed {}", e),
|
279
|
+
}
|
280
|
+
});
|
281
|
+
|
282
|
+
// first rows for schema inference
|
283
|
+
let mut buf = Vec::with_capacity(inference_size);
|
284
|
+
buf.push(first_value);
|
285
|
+
buf.extend((&mut row_iter).take(inference_size).cloned());
|
286
|
+
let schema = rows_to_schema_first_non_null(&buf, Some(50));
|
287
|
+
|
288
|
+
if init_null_count > 0 {
|
289
|
+
// Safety: we know the iterators size
|
290
|
+
let iter = unsafe {
|
291
|
+
(0..init_null_count)
|
292
|
+
.map(|_| &null_row)
|
293
|
+
.chain(buf.iter())
|
294
|
+
.chain(row_iter)
|
295
|
+
.trust_my_length(df.height())
|
296
|
+
};
|
297
|
+
DataFrame::from_rows_iter_and_schema(iter, &schema)
|
298
|
+
} else {
|
299
|
+
// Safety: we know the iterators size
|
300
|
+
let iter = unsafe { buf.iter().chain(row_iter).trust_my_length(df.height()) };
|
301
|
+
DataFrame::from_rows_iter_and_schema(iter, &schema)
|
302
|
+
}
|
303
|
+
}
|
@@ -0,0 +1,253 @@
|
|
1
|
+
pub mod dataframe;
|
2
|
+
pub mod series;
|
3
|
+
|
4
|
+
use magnus::{RHash, Value};
|
5
|
+
use polars::chunked_array::builder::get_list_builder;
|
6
|
+
use polars::prelude::*;
|
7
|
+
use polars_core::export::rayon::prelude::*;
|
8
|
+
use polars_core::utils::CustomIterTools;
|
9
|
+
use polars_core::POOL;
|
10
|
+
|
11
|
+
use crate::{ObjectValue, RbPolarsErr, RbResult, RbSeries, Wrap};
|
12
|
+
|
13
|
+
pub trait RbArrowPrimitiveType: PolarsNumericType {}
|
14
|
+
|
15
|
+
impl RbArrowPrimitiveType for UInt8Type {}
|
16
|
+
impl RbArrowPrimitiveType for UInt16Type {}
|
17
|
+
impl RbArrowPrimitiveType for UInt32Type {}
|
18
|
+
impl RbArrowPrimitiveType for UInt64Type {}
|
19
|
+
impl RbArrowPrimitiveType for Int8Type {}
|
20
|
+
impl RbArrowPrimitiveType for Int16Type {}
|
21
|
+
impl RbArrowPrimitiveType for Int32Type {}
|
22
|
+
impl RbArrowPrimitiveType for Int64Type {}
|
23
|
+
impl RbArrowPrimitiveType for Float32Type {}
|
24
|
+
impl RbArrowPrimitiveType for Float64Type {}
|
25
|
+
|
26
|
+
fn iterator_to_struct(
|
27
|
+
it: impl Iterator<Item = Option<Value>>,
|
28
|
+
init_null_count: usize,
|
29
|
+
first_value: AnyValue,
|
30
|
+
name: &str,
|
31
|
+
capacity: usize,
|
32
|
+
) -> RbResult<RbSeries> {
|
33
|
+
let (vals, flds) = match &first_value {
|
34
|
+
av @ AnyValue::Struct(_, _, flds) => (av._iter_struct_av().collect::<Vec<_>>(), &**flds),
|
35
|
+
AnyValue::StructOwned(payload) => (payload.0.clone(), &*payload.1),
|
36
|
+
_ => {
|
37
|
+
return Err(crate::error::ComputeError::new_err(format!(
|
38
|
+
"expected struct got {first_value:?}",
|
39
|
+
)))
|
40
|
+
}
|
41
|
+
};
|
42
|
+
|
43
|
+
let struct_width = vals.len();
|
44
|
+
|
45
|
+
// every item in the struct is kept as its own buffer of anyvalues
|
46
|
+
// so as struct with 2 items: {a, b}
|
47
|
+
// will have
|
48
|
+
// [
|
49
|
+
// [ a values ]
|
50
|
+
// [ b values ]
|
51
|
+
// ]
|
52
|
+
let mut items = Vec::with_capacity(vals.len());
|
53
|
+
for item in vals {
|
54
|
+
let mut buf = Vec::with_capacity(capacity);
|
55
|
+
for _ in 0..init_null_count {
|
56
|
+
buf.push(AnyValue::Null);
|
57
|
+
}
|
58
|
+
buf.push(item.clone());
|
59
|
+
items.push(buf);
|
60
|
+
}
|
61
|
+
|
62
|
+
for dict in it {
|
63
|
+
match dict {
|
64
|
+
None => {
|
65
|
+
for field_items in &mut items {
|
66
|
+
field_items.push(AnyValue::Null);
|
67
|
+
}
|
68
|
+
}
|
69
|
+
Some(dict) => {
|
70
|
+
let dict = dict.try_convert::<RHash>()?;
|
71
|
+
if dict.len() != struct_width {
|
72
|
+
return Err(crate::error::ComputeError::new_err(
|
73
|
+
format!("Cannot create struct type.\n> The struct dtype expects {} fields, but it got a dict with {} fields.", struct_width, dict.len())
|
74
|
+
));
|
75
|
+
}
|
76
|
+
// we ignore the keys of the rest of the dicts
|
77
|
+
// the first item determines the output name
|
78
|
+
todo!()
|
79
|
+
// for ((_, val), field_items) in dict.iter().zip(&mut items) {
|
80
|
+
// let item = val.try_convert::<Wrap<AnyValue>>()?;
|
81
|
+
// field_items.push(item.0)
|
82
|
+
// }
|
83
|
+
}
|
84
|
+
}
|
85
|
+
}
|
86
|
+
|
87
|
+
let fields = POOL.install(|| {
|
88
|
+
items
|
89
|
+
.par_iter()
|
90
|
+
.zip(flds)
|
91
|
+
.map(|(av, fld)| Series::new(fld.name(), av))
|
92
|
+
.collect::<Vec<_>>()
|
93
|
+
});
|
94
|
+
|
95
|
+
Ok(StructChunked::new(name, &fields)
|
96
|
+
.unwrap()
|
97
|
+
.into_series()
|
98
|
+
.into())
|
99
|
+
}
|
100
|
+
|
101
|
+
fn iterator_to_primitive<T>(
|
102
|
+
it: impl Iterator<Item = Option<T::Native>>,
|
103
|
+
init_null_count: usize,
|
104
|
+
first_value: Option<T::Native>,
|
105
|
+
name: &str,
|
106
|
+
capacity: usize,
|
107
|
+
) -> ChunkedArray<T>
|
108
|
+
where
|
109
|
+
T: RbArrowPrimitiveType,
|
110
|
+
{
|
111
|
+
// safety: we know the iterators len
|
112
|
+
let mut ca: ChunkedArray<T> = unsafe {
|
113
|
+
if init_null_count > 0 {
|
114
|
+
(0..init_null_count)
|
115
|
+
.map(|_| None)
|
116
|
+
.chain(std::iter::once(first_value))
|
117
|
+
.chain(it)
|
118
|
+
.trust_my_length(capacity)
|
119
|
+
.collect_trusted()
|
120
|
+
} else if first_value.is_some() {
|
121
|
+
std::iter::once(first_value)
|
122
|
+
.chain(it)
|
123
|
+
.trust_my_length(capacity)
|
124
|
+
.collect_trusted()
|
125
|
+
} else {
|
126
|
+
it.collect()
|
127
|
+
}
|
128
|
+
};
|
129
|
+
debug_assert_eq!(ca.len(), capacity);
|
130
|
+
ca.rename(name);
|
131
|
+
ca
|
132
|
+
}
|
133
|
+
|
134
|
+
fn iterator_to_bool(
|
135
|
+
it: impl Iterator<Item = Option<bool>>,
|
136
|
+
init_null_count: usize,
|
137
|
+
first_value: Option<bool>,
|
138
|
+
name: &str,
|
139
|
+
capacity: usize,
|
140
|
+
) -> ChunkedArray<BooleanType> {
|
141
|
+
// safety: we know the iterators len
|
142
|
+
let mut ca: BooleanChunked = unsafe {
|
143
|
+
if init_null_count > 0 {
|
144
|
+
(0..init_null_count)
|
145
|
+
.map(|_| None)
|
146
|
+
.chain(std::iter::once(first_value))
|
147
|
+
.chain(it)
|
148
|
+
.trust_my_length(capacity)
|
149
|
+
.collect_trusted()
|
150
|
+
} else if first_value.is_some() {
|
151
|
+
std::iter::once(first_value)
|
152
|
+
.chain(it)
|
153
|
+
.trust_my_length(capacity)
|
154
|
+
.collect_trusted()
|
155
|
+
} else {
|
156
|
+
it.collect()
|
157
|
+
}
|
158
|
+
};
|
159
|
+
debug_assert_eq!(ca.len(), capacity);
|
160
|
+
ca.rename(name);
|
161
|
+
ca
|
162
|
+
}
|
163
|
+
|
164
|
+
fn iterator_to_object(
|
165
|
+
it: impl Iterator<Item = Option<ObjectValue>>,
|
166
|
+
init_null_count: usize,
|
167
|
+
first_value: Option<ObjectValue>,
|
168
|
+
name: &str,
|
169
|
+
capacity: usize,
|
170
|
+
) -> ObjectChunked<ObjectValue> {
|
171
|
+
// safety: we know the iterators len
|
172
|
+
let mut ca: ObjectChunked<ObjectValue> = unsafe {
|
173
|
+
if init_null_count > 0 {
|
174
|
+
(0..init_null_count)
|
175
|
+
.map(|_| None)
|
176
|
+
.chain(std::iter::once(first_value))
|
177
|
+
.chain(it)
|
178
|
+
.trust_my_length(capacity)
|
179
|
+
.collect_trusted()
|
180
|
+
} else if first_value.is_some() {
|
181
|
+
std::iter::once(first_value)
|
182
|
+
.chain(it)
|
183
|
+
.trust_my_length(capacity)
|
184
|
+
.collect_trusted()
|
185
|
+
} else {
|
186
|
+
it.collect()
|
187
|
+
}
|
188
|
+
};
|
189
|
+
debug_assert_eq!(ca.len(), capacity);
|
190
|
+
ca.rename(name);
|
191
|
+
ca
|
192
|
+
}
|
193
|
+
|
194
|
+
fn iterator_to_utf8(
|
195
|
+
it: impl Iterator<Item = Option<String>>,
|
196
|
+
init_null_count: usize,
|
197
|
+
first_value: Option<&str>,
|
198
|
+
name: &str,
|
199
|
+
capacity: usize,
|
200
|
+
) -> Utf8Chunked {
|
201
|
+
let first_value = first_value.map(|v| v.to_string());
|
202
|
+
|
203
|
+
// safety: we know the iterators len
|
204
|
+
let mut ca: Utf8Chunked = unsafe {
|
205
|
+
if init_null_count > 0 {
|
206
|
+
(0..init_null_count)
|
207
|
+
.map(|_| None)
|
208
|
+
.chain(std::iter::once(first_value))
|
209
|
+
.chain(it)
|
210
|
+
.trust_my_length(capacity)
|
211
|
+
.collect_trusted()
|
212
|
+
} else if first_value.is_some() {
|
213
|
+
std::iter::once(first_value)
|
214
|
+
.chain(it)
|
215
|
+
.trust_my_length(capacity)
|
216
|
+
.collect_trusted()
|
217
|
+
} else {
|
218
|
+
it.collect()
|
219
|
+
}
|
220
|
+
};
|
221
|
+
debug_assert_eq!(ca.len(), capacity);
|
222
|
+
ca.rename(name);
|
223
|
+
ca
|
224
|
+
}
|
225
|
+
|
226
|
+
fn iterator_to_list(
|
227
|
+
dt: &DataType,
|
228
|
+
it: impl Iterator<Item = Option<Series>>,
|
229
|
+
init_null_count: usize,
|
230
|
+
first_value: Option<&Series>,
|
231
|
+
name: &str,
|
232
|
+
capacity: usize,
|
233
|
+
) -> RbResult<ListChunked> {
|
234
|
+
let mut builder =
|
235
|
+
get_list_builder(dt, capacity * 5, capacity, name).map_err(RbPolarsErr::from)?;
|
236
|
+
for _ in 0..init_null_count {
|
237
|
+
builder.append_null()
|
238
|
+
}
|
239
|
+
builder.append_opt_series(first_value);
|
240
|
+
for opt_val in it {
|
241
|
+
match opt_val {
|
242
|
+
None => builder.append_null(),
|
243
|
+
Some(s) => {
|
244
|
+
if s.len() == 0 && s.dtype() != dt {
|
245
|
+
builder.append_series(&Series::full_null("", 0, dt))
|
246
|
+
} else {
|
247
|
+
builder.append_series(&s)
|
248
|
+
}
|
249
|
+
}
|
250
|
+
}
|
251
|
+
}
|
252
|
+
Ok(builder.finish())
|
253
|
+
}
|