forked from extern/nushell
9ce61dc677
Closes #9003. This PR changes `group-by` so that its optional argument is interpreted as a cell path. In turn, this lets users use `?` to ignore rows that are missing the column they wish to group on. For example: ``` > [{foo: 123}, {foo: 234}, {bar: 345}] | group-by foo Error: nu:🐚:column_not_found × Cannot find column ╭─[entry #3:1:1] 1 │ [{foo: 123}, {foo: 234}, {bar: 345}] | group-by foo · ─────┬──── ─┬─ · │ ╰── cannot find column 'foo' · ╰── value originates here ╰──── > [{foo: 123}, {foo: 234}, {bar: 345}] | group-by foo? ╭─────┬───────────────╮ │ 123 │ [table 1 row] │ │ 234 │ [table 1 row] │ ╰─────┴───────────────╯ ``` ~~This removes the ability to pass `group-by` a closure or block (I wasn't able to figure out how to make the 2 features coexist), and so it is a breaking change. I think this is OK; I didn't even know `group-by` could accept a closure or block because there was no example for that functionality.~~
111 lines
3.2 KiB
Rust
111 lines
3.2 KiB
Rust
use nu_test_support::fs::Stub::FileWithContentToBeTrimmed;
|
|
use nu_test_support::playground::Playground;
|
|
use nu_test_support::{nu, pipeline};
|
|
|
|
#[test]
|
|
fn groups() {
|
|
Playground::setup("group_by_test_1", |dirs, sandbox| {
|
|
sandbox.with_files(vec![FileWithContentToBeTrimmed(
|
|
"los_tres_caballeros.csv",
|
|
r#"
|
|
first_name,last_name,rusty_at,type
|
|
Andrés,Robalino,10/11/2013,A
|
|
JT,Turner,10/12/2013,B
|
|
Yehuda,Katz,10/11/2013,A
|
|
"#,
|
|
)]);
|
|
|
|
let actual = nu!(
|
|
cwd: dirs.test(), pipeline(
|
|
r#"
|
|
open los_tres_caballeros.csv
|
|
| group-by rusty_at
|
|
| get "10/11/2013"
|
|
| length
|
|
"#
|
|
));
|
|
|
|
assert_eq!(actual.out, "2");
|
|
})
|
|
}
|
|
|
|
#[test]
|
|
fn errors_if_given_unknown_column_name() {
|
|
Playground::setup("group_by_test_2", |dirs, sandbox| {
|
|
sandbox.with_files(vec![FileWithContentToBeTrimmed(
|
|
"los_tres_caballeros.json",
|
|
r#"
|
|
{
|
|
"nu": {
|
|
"committers": [
|
|
{"name": "Andrés N. Robalino"},
|
|
{"name": "JT Turner"},
|
|
{"name": "Yehuda Katz"}
|
|
],
|
|
"releases": [
|
|
{"version": "0.2"}
|
|
{"version": "0.8"},
|
|
{"version": "0.9999999"}
|
|
],
|
|
"0xATYKARNU": [
|
|
["Th", "e", " "],
|
|
["BIG", " ", "UnO"],
|
|
["punto", "cero"]
|
|
]
|
|
}
|
|
}
|
|
"#,
|
|
)]);
|
|
|
|
let actual = nu!(
|
|
cwd: dirs.test(), pipeline(
|
|
r#"
|
|
open los_tres_caballeros.json
|
|
| group-by {|| get nu.releases.version }
|
|
"#
|
|
));
|
|
|
|
assert!(actual
|
|
.err
|
|
.contains("requires a table with one value for grouping"));
|
|
})
|
|
}
|
|
|
|
#[test]
|
|
fn errors_if_column_not_found() {
|
|
Playground::setup("group_by_test_3", |dirs, sandbox| {
|
|
sandbox.with_files(vec![FileWithContentToBeTrimmed(
|
|
"los_tres_caballeros.csv",
|
|
r#"
|
|
first_name,last_name,rusty_at,type
|
|
Andrés,Robalino,10/11/2013,A
|
|
JT,Turner,10/12/2013,B
|
|
Yehuda,Katz,10/11/2013,A
|
|
"#,
|
|
)]);
|
|
|
|
let actual = nu!(
|
|
cwd: dirs.test(), pipeline(
|
|
r#"
|
|
open los_tres_caballeros.csv
|
|
| group-by ttype
|
|
"#
|
|
));
|
|
|
|
assert!(actual.err.contains("did you mean 'type'"),);
|
|
})
|
|
}
|
|
|
|
#[test]
|
|
fn errors_if_input_empty() {
|
|
let actual = nu!("group-by date");
|
|
assert!(actual.err.contains("expected table from pipeline"));
|
|
}
|
|
|
|
#[test]
|
|
fn optional_cell_path_works() {
|
|
let actual = nu!("[{foo: 123}, {foo: 234}, {bar: 345}] | group-by foo? | to nuon");
|
|
let expected = r#"{"123": [[foo]; [123]], "234": [[foo]; [234]]}"#;
|
|
assert_eq!(actual.out, expected)
|
|
}
|