NEWS

---

censobr 1.0.0 (2026-09-21)

• New data release v1.0.0, which includes the following news files or edits:

  • Census 2022. Closes #65
    • Data dictionary of microdata
    • Public microdata.
    • New function import_microdata22(), which brings the controlled-access microdata of the 2022 Population Census into censobr. See new features below.
  • Data fixes
    • 2022: census tract, table "preliminares"
    • 2010: census tract, table "pessoas", state of Sao Paulo
    • 1980: code_muni values of Tocantins and Fernando de Noronha
    • 1970: variables weight_household and hh_income are stored as integers. Totals weighted by weight_household therefore differ slightly from earlier releases -- sum(weight_household) goes from 17,682,112 to 17,643,387 (-0.22%).
    • 1960. Major updates. See details in v1.0.0.
    • All year: the codes of categorical variables of the microdata are stored as integers rather than text, in every census year. Code that filtered on the text form needs to drop the quotes, e.g. filter(V0601 == "1") becomes filter(V0601 == 1). Zero-padded codes lose the padding as well: V0402 == "01" becomes V0402 == 1. Values that used to be recorded as "." are now NA.

• New features

  • New function import_microdata22() to import to censobr the controlled-access microdata of 2022. Once the .zip file with the original data is imported read_(year = 2022) functions always read the controlled-access microdata. If the controlled-access microdata have not been imported yet, these functions return an informative warning and download the public microdata set, which has fewer variables. See the new vignette Working with 2022 microdata. Closes #79.

• Major changes

  • add_labels = "pt" now works for all years and tables since 1960. Closes #25, #26, and #27.
  • merge_households parameter now works for all census years since 1960. Because merging all ~300 population + househols columns can require more than 20GB of memory, read_population(merge_households = TRUE) requires columns to be set -- naming the columns you need keeps the operation to a few seconds and a few dozen MB. Closes #31.
  • data_dictionary() now takes only two values in dataset: "microdata", which opens a single Excel file covering every variable of the microdata and is now available for all censuses since 1960, and "tracts", available since 1970.
  • Microdata dictionary now includes in a single file all supplementary dictionaries of variable codes (e.g. occupation, religion, education categories etc). Closes 53.

• Breaking changes

  • All functions that take a year now require the user to declare it. Previously questionnaire() silently assumed year = 2010.
  • The arguments questionnaire(type), read_tracts(dataset) and data_dictionary(dataset) are now explicitly required, and the error message lists the values accepted.
  • data_dictionary() now takes only two values in dataset: "microdata", which opens a single Excel file covering every variable of the microdata. The per-data-set dictionaries opened with dataset = "population" and dataset = "households" were retired, since the microdata dictionary now covers the pre-2000 censuses too.

• Minor changes

  • censobr now uses {httr2} to download files, replacing {curl}.
  • data_dictionary(), questionnaire() and interview_manual() now return the path to the downloaded file. The file is only opened when verbose = TRUE and the session is interactive, so scripted runs no longer launch a viewer.
  • columns now only accepts a character vector of column names, in all five microdata readers (read_population(), read_households(), read_families(), read_mortality(), read_emigration()), matching its documented type. It previously also silently accepted numeric column indices.
  • The argument dataset in data_dictionary() is also case insensitive now, as in read_tracts().

• bug fixes

  • Several bug fixes and a few label corrections when add_labels = "pt" in multiple read_ functions.
  • add_labels = "pt" now compares the codes of every census year as numbers, matching how the microdata are stored since the v0.7.0 data release.
  • Requesting a column that does not exist now returns an informative error naming the column, instead of an internal {dplyr} message.
  • Passing more than one year now returns an informative error. Previously a vector such as year = c(2000, 2010) failed with a cryptic "the condition has length > 1" message from base R.
  • An incomplete download is now detected by comparing the size of the file with the size reported by the server, and is removed instead of being cached.
  • A corrupted file in the local cache no longer throws an error. The file is removed and the function returns NULL, so that running it again downloads a fresh copy instead of failing on every call.
  • Passing cache = FALSE no longer fails when the cache directory does not exist yet, for example on a fresh installation.
  • Download error messages now match their cause. A failed transfer no longer reports the local file as corrupted, and an incomplete download no longer reports the internet connection as faulty.
  • The temporary DuckDB database file created by merge_households = TRUE is now removed when the merge finishes. Previously it was left behind in the session's temp directory.
  • Data cached from previous data releases is deleted again. The cache directory is versioned by data release, so a new release leaves the files of the previous one behind. censobr used to delete them when the package was loaded, but that stopped working in v0.6.0, when the cache directory became a setting the user can change. The check now runs once per session, the first time data is downloaded, and can also be run on demand with censobr_cache(delete_file = 'old'). Set options(censobr.keep_old_cache = TRUE) to keep those files, for example to go on working with an older data release. Microdata imported with import_microdata22() are never deleted, because censobr cannot download them again.

censobr 0.6.0 (2026-07-29)

• Minor changes

  • The function data_dictionary() now does not open the file when verbose = FALSE. Closes 72
  • All data, documentation files and pipeline to generate the data sets shared through the censobr package have been migrated repository to https://github.com/ipea/censobr_prep_data.

• Data fixes included in this version:

  • The census tract aggregate table of Pessoa02 from the state of Goias has been fixed. Closes 68, 70 and 71.
  • All code_ columns now have class numeric to keep the consistency across {geobr} and other sister packages in the brverse.

• New data set and files included in this version:

  • Data dictionary of microdata now includes a single Excel file with info for all variables, including auxiliary documentation. For now, available for the years 2000 and 2010.

censobr 0.5.0 (2025-07-07)

• Major changes

  • New function get_censobr_cache_dir()
  • The function set_censobr_cache_dir() now sets cache directories that persist across R sessions. Closes #55. The data is saved in versioned directory inside the cache directory.
  • The year parameter no longer defaults to 2010.
  • New parameter verbose (logical) indicating whether functions should print messsages

• Minor changes

  • Improved internal code of merge_households = TRUE to avoid duplicated columns
  • Improved package info and error messages with {cli}
  • {censobr} now imports {cli} and {rlang}

• New data set and files included in this version:

  • 2022 census. Closes #64
    • Census-tract level data
    • Census-tract level data dictionary
  • 2000 census. Closes #43
    • Census-tract level data
  • All data sets are save in .parquet compressed using compression='zstd' and compression_level = 22. This has almost halved the size of data files, making downloads much more efficient at minimal cost of reading time.
  • All data sets are now sorted by key columns to speed up join operations. Closes #60.
  • Fixed annoying message about arrow metadata. closed #56.

censobr 0.4.1 (2025-02-28)

• Minor changes

  • Removed {duckplyr} from package dependency

• bug fixes

  • Passing parameter merge_households = TRUE now returns the expected result.

censobr 0.4.0 (2024-09-20)

• Major changes

  • Some functions (read_mortality, read_emigration) now include a new parameter merge_households (logical) to indicate whether the function should merge household variables to the output data. Partially closes #31
  • {censobr} now imports the {duckplyr} package, which is used for merging household data. Closes issue #31.
  • New vignette showing how to work with larger-than-memory data. Closes #42. The vignette still needs to be expanded with more examples, though.

• Minor changes

  • Updated Vignettes Closes issue #51
  • Removed dependency on the {httr} package
  • Now using curl::multi_download() to download files in parallel. This brings the advantage that the package now automatically detects whether the data/documentation file has been updated and should be downloaded again.

• Changes to data sets and files included in this version:

  • Population microdata for the year 2000 now include a few columns that were not included before. Closes #44
  • Included additional columns and fixed minor errors in data dictionary of 2010 microdata. Closes #45

• New data set and files included in this version:

  • 1960 census Closes #32
    • Interview manual
    • Data dictionary for microdata of population and households
    • Microdata of population and households
  • 1970: fixed geography columns. Closes #52
  • 1991 census: Data dictionary for microdata of population and households. Closes #28

censobr 0.3.2 (2024-03-23)

• Minor changes

  • Moved {arrow} package back to Imports

• New data set and files included in this version:

  • 2022 census
    • Preliminary aggregate results of census tracts

censobr 0.3.1 (2024-02-23)

• Minor changes
  • Moved {arrow} package from Imports to Suggests while the {arrow} team fixes their conflict with CRAN policies related to downloading binary software. See here.
• New package contributors:
  • Diego Rabatone Oliveira
  • Neal Richardson

censobr 0.3.0 (2024-01-09)

• Major changes

  • The questionnaire() function now accepts questionnaires of type: "long" or "short".
  • Updated census tract data following latest update by IBGE on Oct/2023. Closed #38. As a result, the package moved to data release v0.3.0.

• Minor changes

  • Replaced .onAttach by .onLoad so that the package works with censobr::function()
  • Fixed documentation of various functions.
  • Fixed issue to make sure censobr uses suggested packages conditionally on CRAN
  • Fixed message when user requests a data set / file for a year that is not available

• New data set and files included in this version:

  • 2022 census [New]
    • Questionnaires and interview manuals
  • Short questionnaires for every census between 1960 and 2022.
  • Long questionnaire for the 1960 and 2022 censuses.

censobr 0.2.0 (2023-09-30)

• Major changes

  • New function read_tracts() to read Census tract-level aggregate data.
  • New function data_dictionary() opens on a browser the data dictionary of Brazil's census data.
  • New function questionnaire() opens on a browser the questionnaire used in the data collection of Brazil's censuses.
  • New function interview_manual() opens on a browser the interview manual of the data collection of Brazil's censuses.
    • New function set_censobr_cache_dir() that allows users to set custom directory for caching files from the censobr package.
  • New data sets of 1970, 1980 and 1991 censuses: microdata of population and households PLUS Census tract-level aggregate data for 2010. Closes #6, #7, #8 and 1#8
  • New vignette on Census tract-level aggregate data for 2010.
  • New vignette covering functions about census documentation and dictionary of variables. Closes #2.

• Minor changes

  • Running censobr_cache(delete_file = 'all') now removes all data and directories related from censobr.
  • censobr now uses suggested package {geobr} conditionally

• Data included in this version:

  • 1970 census [New]
    • Microdata of population, households
  • 1980 census [New]
    • Microdata of population, households
  • 1991 census [New]
    • Microdata of population, households
  • 2000 Census
    • Microdata of population, households and families.
  • 2010 Census
    • Microdata of population, households, deaths and emigration.
    • Census tract-level aggregate data [New]

censobr 0.1.1 (2023-09-11)

• Minor changes

  • Using cache_dir and data_release as global variables. Closes #13
  • Running censobr_cache(delete_file = 'all')now also remove data from old data releases. Closes #14.
  • Large improvement in code coverage

• Changes requested by CRAN team

  • Changed location of cached data to directory inside tools::R_user_dir("censobr", which = "cache").
  • The package now automatically deletes cached data from previous data releases that might exist from previous versions of the package
  • Clean cache after intro vignette and testhat checks

censobr 0.1.0 (2023-09-06)

• Launch of censobr v0.1.0 on CRAN https://cran.r-project.org/package=censobr
• All data sets are now enriched with geography columns following {geobr} name standards. This should help data manipulation and integration with spatial data from the {geobr} package. The added columns are: c('code_muni', 'code_state', 'abbrev_state', 'name_state', 'code_region', 'name_region', 'code_weighting'). Closes #5.
• Data included in this version:
  • 2000 Census
    • Microdata of population, households and families.
  • 2010 Census
    • Microdata of population, households, deaths and emigration.