forked from extern/nushell
0569a9c92e
# Description Pretty much all operations/commands in Nushell assume that the column names/keys in a record and thus also in a table (which consists of a list of records) are unique. Access through a string-like cell path should refer to a single column or key/value pair and our output through `table` will only show the last mention of a repeated column name. ```nu [[a a]; [1 2]] ╭─#─┬─a─╮ │ 0 │ 2 │ ╰───┴───╯ ``` While the record parsing already either errors with the `ShellError::ColumnDefinedTwice` or silently overwrites the first occurence with the second occurence, the table literal syntax `[[header columns]; [val1 val2]]` currently still allowed the creation of tables (and internally records with more than one entry with the same name. This is not only confusing, but also breaks some assumptions around how we can efficiently perform operations or in the past lead to outright bugs (e.g. #8431 fixed by #8446). This PR proposes to make this an error. After this change another hole which allowed the construction of records with non-unique column names will be plugged. ## Parts - Fix `SE::ColumnDefinedTwice` error code - Remove previous tests permitting duplicate columns - Deny duplicate column in table literal eval - Deny duplicate column in const eval - Deny duplicate column in `from nuon` # User-Facing Changes `[[a a]; [1 2]]` will now return an error: ``` Error: nu:🐚:column_defined_twice × Record field or table column used twice ╭─[entry #2:1:1] 1 │ [[a a]; [1 2]] · ┬ ┬ · │ ╰── field redefined here · ╰── field first defined here ╰──── ``` this may under rare circumstances block code from evaluating. Furthermore this makes some NUON files invalid if they previously contained tables with repeated column names. # Tests + Formatting Added tests for each of the different evaluation paths that materialize tables. |
||
|---|---|---|
| .. | ||
| src | ||
| tests | ||
| Cargo.toml | ||
| LICENSE | ||