Fixes #463 Exposure (or lack thereof) of syntactic argument in tibble() and as_tibble() Fixes #461 Advertise .tidy_names Fixed #446 Justify tidy_names --- Other name-related tibble issues * Finalize interface of `.name_repair = <A_FUNCTION>` #476 *let's decide/discuss there* * tibble doesn't make it easy to get variable names that match our style guide #472 *let's solve in follow-up PR* * names<- method should protect against invalid names #466 *let's solve in follow-up PR* * Move name repair to vctrs #464 *PR lays groundwork for this long-term goal* * revisit strategy for making names syntactic #459 *let's solve in follow-up PR* * Re-encode character columns and column names to UTF-8 #87 *relevant, but I don't touch this* Related issue in [tidyverse/principles](https://github.com/tidyverse/principles): * [Vector names](tidyverse/design#4) --- Overview: * Defines and motivates 3 levels of repaired names: - `minimal`: not `NULL`, no `NA`s, à la `rlang::names2()` - `unique`: `minimal` + no `""`s and no dupes (cf `tidy_names(syntactic = FALSE)`) - `syntactic`: `unique` + syntactic (cf `tidy_names(syntactic = TRUE)`) * Internal utility functions for each level of name repair, i.e. all combinations of "repair vs check" and "operate on the names vs the named vector". * `tibble()` and `as_tibble()` expose name repair via `.name_repair`. Replaces `.tidy_names` (piloted only in dev) and `as_tibble(..., validate = TRUE/FALSE)` (in CRAN version). * Default behaviour (`.name_repair = "assert_unique")`): do no repair but check that the names are `unique`. * `.name_repair = "none"` means just make sure the names are `minimal` * `.name_repair = "unique"` and `.name_repair = "syntactic"` mean what you think * `.name_repair` can also take a user-supplied function. --- To discuss: * ~Should `valid` just be `unique`?~ YES and I have made it so in the source. * Let us review the "names degrees of freedom" and facilities offered by base, tidyverse for independent control: - exist or not - unique or not - syntactic or not * Automatic naming, based on expression. We currently offer no control over this in `tibble()`. * How much to reveal re: long-term plans. For example, why `tidy_names()` and `set_tidy_names()` are still here, but haven't gained the obvious counterparts like `unique_names()` and `set_syntactic_names()`. Do we need a Life Cycle section in the docs? * My backwards compatibility report

Move name repair to vctrs #464

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions