Skip to contents

Changelog

Source: NEWS.md

airtable2 0.1.2.9000

Breaking changes

  • air_write() and other upload functions now clear values from fields when the corresponding column contains NA or NULL. Previously, these values were ignored, leaving the existing field value unchanged.

  • air_write(), air_upsert(), and air_sync() now error by default when data contains complex list-columns (nested lists or data frames that Airtable cannot store directly). Previously these columns were silently problematic. Set complex_fields = "warn" to drop them or complex_fields = "json" to serialize them as JSON text.

  • add_fields = "yes" in air_write(), air_upsert(), and air_sync() now infers field types from column class instead of always creating singleLineText. Mapping: numeric → number, logical → checkbox, Date → date, complex/JSON columns → multilineText, everything else → singleLineText. This changes which Airtable field type is created when a new column is first written.

New features

  • air_write() gains a create_table argument. When TRUE, the table is created in the base if it does not already exist; field types are inferred from the data using the same rules as add_fields = "yes".

  • air_infer_fields(data) maps R column types to Airtable field specification lists suitable for at_create_table() / at_create_base(). Factors become singleSelect fields with choices from levels().

  • air_infer_table(data, name) wraps air_infer_fields() to produce a complete table specification in one call.

  • air_left_join_upload() gains typecast, complex_fields, and progress arguments, bringing its API in line with air_upsert().

  • When add_fields = "yes" creates multiple new fields, a progress bar is shown instead of one message per field.

Bug fixes

  • air_sync() no longer reports unchanged rows as changed (#13). Change detection now reduces each field to a canonical, type-aware representation before hashing, so a local column compares equal to the value Airtable actually stores and returns. Previously the hash relied on format() happening to agree across representations, which produced spurious updates on every sync for several field types:
    • dateTime — compared the wrong moment (the hash used the absolute UTC instant, but Airtable round-trips datetimes by wall-clock).
    • checkbox — unchecked boxes read back as NA (Airtable omits them) and never matched a local FALSE.
    • multipleSelects / linked records and other list-columns — were padded by format() so a list never matched the identical list read back.
    • number — integer-vs-double and scientific-notation drift (e.g. 1e+05 vs 100000). These are covered by live tests exercising every writable field type.
  • air_restore() now correctly handles tables containing multiselect, linked-record, attachment, or other list-type columns. Previously air_write() would abort with a “complex columns” error because air_dump() stores those columns as plain lists (without the air_* class); they are now re-wrapped before writing.

airtable2 0.0.0.9000

Initial development release — complete rewrite of airtabler using modern R packages.

DBI improvements (Stage 8A)

  • Added dbGetRowsAffected() for AirtableResult — returns NA_integer_ (read-only results satisfy the DBI contract).
  • Added dbFetch() dispatch for missing n so dbFetch(res) works without specifying n.
  • Extended DBI test coverage: dbGetInfo, dbWriteTable modes, dbExistsTable FALSE case, dbListFields error path, dbi_parse_statement edge cases, multi-base dbListTables, dbSendQuery formula forwarding.
  • Added tests/testthat/test-dbitest.R as DBItest conformance placeholder.
  • Added DBI vignette vignettes/dbi.Rmd (Stage 8C).

Breaking changes

  • Normalized function signatures so the piped/main object comes first and the container hierarchy (table → base → workspace) comes last, letting a session default base be used positionally:
    • air_read(table, base_id = NULL, ...) (was base-first)
    • air_write(data, table, base_id = NULL, ...), air_upsert(data, table, merge_on, base_id = NULL, ...), air_sync(data, table, key, base_id = NULL, ...)
    • air_delete(record_ids, table, base_id = NULL, ...) (now resolves the session default base)
    • air_left_join() / air_inner_join() / air_full_join() are now (x, table, base_id = NULL, by = NULL, ...)
    • at_create_base(name, tables, workspace_id = NULL, ...)

Bug fixes

  • Attachment uploads now work. at_upload_attachment() (and the air_write_attachments() / air_sync_attachments() / air_demo() paths that call it) previously POSTed to api.airtable.com and always got a 404, because Airtable serves the uploadAttachment endpoint from a separate host. Uploads now go to content.airtable.com. air_req() gained a host argument to select the host.
  • API errors are now informative: air_perform() surfaces Airtable’s error type and message plus actionable hints (e.g. a wrong table name / missing permission, an unknown field name, or an invalid filterByFormula). This also fixes a dead error handler — the old code caught the non-existent httr2_http_error condition class, so Airtable’s explanation was silently dropped and users only saw the bare HTTP status.
  • air_api_usage() now counts parallel attachment downloads (which previously bypassed the counter via req_perform_parallel()), and its print method shows the workspace ID as a clickable link.
  • Airtable IDs in air_browse() messages are now clickable cli hyperlinks.
  • air_browse() now honors its base_id/table_id arguments (they previously leaked into utils::browseURL()), so browsing a table/view by ID works.
  • .onAttach now warns (once, via packageStartupMessage()) when both airtabler and airtable2 are loaded and their function names would mask.
  • Fixed the pkgdown reference index so the site builds (documented the AirtableDriver topic; indexed air_req()/air_token() and the package doc).

New features

  • Full httr2-based Airtable REST client replacing legacy airtabler (httr-based)
  • Type-aware air_read() / air_write() with S3 list-columns for multiselect, linked records, attachments, collaborators, and barcodes
  • Diff-based air_sync() and air_upsert() that exclude computed/read-only fields automatically using the table schema
  • DBI S4 interface (AirtableDriver, AirtableConnection) for standard database workflows and RStudio/Positron connection pane integration
  • Join helpers: air_left_join(), air_inner_join(), air_full_join()
  • Parallel attachment downloads/uploads (air_read_attachments(), air_write_attachments())
  • Base metadata management: air_meta(), air_meta_push(), air_meta_sync()
  • Schema inspection: air_schema() returns a tidy tibble of tables/fields/types
  • Backup/restore: air_dump() (JSON/CSV + optional attachments) and air_restore()
  • Type helpers: air_flatten_multiselect(), air_flatten_links(), air_flatten_attachments(), air_flatten_collaborator(), air_expand_multiselect(), air_expand_collaborator(), air_simplify()
  • Field/table template builders: air_field_template(), air_table_template()
  • Project-level defaults: air_set_token(), air_set_base() with env var fallback (AIRTABLE_API_KEY, AIRTABLE_BASE_ID)
  • API call tracking: air_api_usage(), respects free-tier ~1000 calls/month limit
  • Situation report: at_sitrep() for debugging credentials and accessible bases
  • Browser navigation: air_browse() to open bases, tables, views, and records
  • Low-level at_* wrappers covering all Airtable REST endpoints (records, tables, bases, fields, views)
  • Convenience wrappers air_connect() / air_pane() for DBI connection pane integration
  • air_delete() high-level record deletion with messaging
  • The default delimiter for flattening/expanding multiselect, links, and attachment fields is now "; " (was ", "), overridable with the airtable2.delimiter option or AIRTABLE2_DELIMITER env var; expansion tolerates spaces around the separator so flatten/expand round-trips
  • air_browse() accepts base/table NAMES (not just IDs), opens the session default base when called with no argument, and resolves app/wsp/tbl/viw prefixes
  • air_attachment_preview_url() plus documentation of Airtable attachment URL behavior (the API url expires ~2h; stable viewer links are obtained in the web app)
  • air_flatten() is now an S3 generic dispatching on the air_multiselect, air_links, air_attachments, and air_collaborator column classes (the per-type air_flatten_*() functions remain as exported wrappers)
  • Write/upsert/sync now auto-expand flat character columns to the structure Airtable expects, inferred from the table schema (e.g. a "; "-joined string into a multiselect or linked-record array, an email into a collaborator object) — manual air_expand_*() is no longer required
  • air_left_join_upload() (was air_upload_join()) joins a local data frame to matched Airtable records and upserts only new or changed field values; reads the remote table minimally (key + existing target fields) and skips unmatched rows without creating records — the upload complement to air_left_join()
  • air_meta_sync() now accepts a local data frame, a CSV/JSON file path, or a table name (defaults to "_metadata") and performs a pull-then-patch: it fetches the current remote metadata, computes a diff, and pushes only changed fields
  • air_meta_init() seeds a _metadata table in a base from a schema data frame or an existing schema object, creating the table and populating it in one call
  • Added vignette("dbi", package = "airtable2") — “Using airtable2 with DBI”: covers connecting via air_connect() / DBI::dbConnect(), exploring connections, reading (full table and formula-filtered queries), writing (append and overwrite/sync), multi-base mode, and DBI limitations