Read JMP (.jmp) tables directly in R, without Python or
a JMP installation. ## Install
install.packages("remotes")
remotes::install_github("JosephHill04/jmpread")For an installation archive supplied separately:
install.packages("jmpread_0.1.0.tar.gz", repos = NULL, type = "source")Windows users can install the supplied .zip with
repos = NULL and type = "win.binary". No
additional runtime R packages or compiled code are required.
All distributed example data comes from the MIT-licensed JMPReader.jl fixtures.
library(jmpread)
path <- system.file("extdata", "jmpreader_jl", "example1", "example1.jmp",
package = "jmpread")
table <- read_jmp(path)
table[, c("ints", "floats")]
# Select columns, preserving their order in the source table.
read_jmp(path, select = c("floats", "ints"))
# Return ordinary vectors without source metadata attributes.
read_jmp(path, metadata = FALSE)For data you are authorised to use, replace path with
your own local file path. Do not upload private or unlicensed data in an
issue or contribution.
JMP temporal values use seconds since 1904-01-01. Date-looking values
can contain time of day, so date columns use POSIXct to retain
precision. Use temporal = "raw" for original seconds. Empty
strings remain ""; explicit pooled missing values become
NA_character_. Numeric missing sentinels become NA.
Formulas remain in raw metadata and are never executed.
Tests cover 17 MIT-licensed JMPReader.jl fixtures, including numeric, text, missing, compressed and temporal values. Comparisons against the pinned Python reader agree across 168,717 cells. A separate comparison against an original Excel source imported into JMP Pro 17.0.0 agrees exactly across 890,336 shared cells, including 38,374 missing values. The private comparison data is not distributed. See validation results for scope.
Use ?read_jmp for the function reference and
vignette("getting-started", package = "jmpread") for
examples. Report problems through Issues with
your R version, package version and a small example you can share.
Compatibility outside the documented and tested encodings is
unconfirmed. Notes, units, labels and formulas are retained as raw
blocks, not structured fields. Row-state flags and palette meanings are
not interpreted. Embedded NUL text is rejected. The full file is read
into memory, and raw metadata increases memory use. Default input and
declared per-column inflation limits are 256 MiB; they are not
protection against adversarial compressed files. See
?read_jmp.
MIT. Format work and algorithms are adapted from jumpio, pinned to
e9cf1b85016637f90cbfca0760ded736475d20c1, and JMPReader.jl.
Upstream notices and fixture provenance are included in inst/. This
project is independent of JMP and its owner.
Maintainer: Joseph Hill joe@sixhillsgroup.com.au.