datamash

TLDR

Get max, min, mean and median of a single column of numbers

$ seq 3 | datamash max 1 min 1 mean 1 median 1

Get the mean of a single column of float numbers (floats must use "," and not ".")

$ echo -e '1.0\n2.5\n3.1\n4.3\n5.6\n5.7' | tr '.' ',' | datamash mean 1

Get the mean of a single column of numbers with a given decimal precision

$ echo -e '1\n2\n3\n4\n5\n5' | datamash -R [number_of_decimals_wanted] mean 1

Get the mean of a single column of numbers ignoring "Na" and "NaN" (literal) strings

$ echo -e '1\n2\nNa\n3\nNaN' | datamash --narm mean 1

SYNOPSIS

datamash [OPTION] op [fld] [op fld ...]

DESCRIPTION

Performs numeric/string operations on input from stdin.

'op' is the operation to perform. If a primary operation is used, it must be listed first, optionally followed by other operations. 'fld' is the input field to use. 'fld' can be a number (1=first field), or a field name when using the -H or --header-in options. Multiple fields can be listed with a comma (e.g. 1,6,8). A range of fields can be listed with a dash (e.g. 2-8). Use colons for operations which require a pair of fields (e.g. 'pcov 2:6').

Primary operations:

groupby, crosstab, transpose, reverse, check

Line-Filtering operations:

rmdup

Per-Line operations:

base64, debase64, md5, sha1, sha224, sha256, sha384, sha512, bin, strbin, round, floor, ceil, trunc, frac, dirname, basename, barename, extname, getnum, cut

Numeric Grouping operations:

sum, min, max, absmin, absmax, range

Textual/Numeric Grouping operations:

count, first, last, rand, unique, collapse, countunique

Statistical Grouping operations:

mean, geomean, harmmean, trimmean, median, q1, q3, iqr, perc, mode, antimode, pstdev, sstdev, pvar, svar, mad, madraw, pskew, sskew, pkurt, skurt, dpo, jarque, scov, pcov, spearson, ppearson

Grouping Options:

-C, --skip-comments: skip comment lines (starting with '#' or ';' and optional whitespace)
-f, --full: print entire input line before op results (default: print only the grouped keys)
-g, --group=X[,Y,Z]: group via fields X,[Y,Z]; equivalent to primary operation 'groupby'
--header-in: first input line is column headers
--header-out: print column headers as first line
-H, --headers: same as '--header-in --header-out'
-i, --ignore-case: ignore upper/lower case when comparing text; this affects grouping, and string operations
-s, --sort: sort the input before grouping; this removes the need to manually pipe the input through 'sort'

File Operation Options:

--no-strict: allow lines with varying number of fields
--filler=X: fill missing values with X (default %s)

General Options:

-t, --field-separator=X: use X instead of TAB as field delimiter
--format=FORMAT: print numeric values with printf style floating-point FORMAT.
--output-delimiter=X: use X instead as output field delimiter (default: use same delimiter as -t/-W)
--narm: skip NA/NaN values
-R, --round=N: round numeric output to N decimal places
-W, --whitespace: use whitespace (one or more spaces and/or tabs) for field delimiters
-z, --zero-terminated: end lines with 0 byte, not newline
--help: display this help and exit
--version: output version information and exit

OPTIONS

AVAILABLE OPERATIONS

Primary Operations

Primary operations affect the way the file is processed. If used, the primary operation must be listed first. Some operations require field numbers (groupby, crosstab) while others do not (reverse,check,transpose). If primary operation is not listed the entire file is processed - either line-by-line (for 'per-line' operations) or all lines as one group (for grouping operations). See Examples section below.

groupby X,Y,... op fld ...: group the file by given fields. Equivalent to option '-g'. For each group perform operation op on field fld.
crosstab X,Y [op fld ...]: cross-tabulate a file by two fields (cross-tabulation is also known as pivot tables). If no operation is specified, counts how many incidents exist of X,Y.
transpose: transpose rows, columns of the input file
reverse: reverse field order in each line
check [N lines] [N fields]: verify the input file has same number of fields in all lines, or the expected number of lines/fields. number of lines and fields are printed to STDOUT. Exits with non-zero code and prints the offending line if there's a mismatch in the number of lines/ fields.

Line-Filtering operations

rmdup: remove lines with duplicated key value

Per-Line operations

base64: Encode the field as base64
debase64: Decode the field as base64, exit with error if invalid base64 string
md5/sha1/sha224/sha256/sha384/sha512: Calculate md5/sha1/sha224/sha256/sha384/sha512 hash of the field value
bin[:BUCKET-SIZE]: bin numeric values into buckets of size BUCKET-SIZE (defaults to 100).
strbin[:BUCKET-SIZE]: hashes the input and returns a numeric integer value between zero and BUCKET-SIZE (defaults to 10).
round/floor/ceil/trunc/frac: numeric rounding operations. round (round half away from zero), floor (round up), ceil (ceiling, round down), trunc (truncate, round towards zero), frac (fraction, return fraction part of a decimal-point value).
dirname/basename: extract the directory name and the base file name from a given string (same as to dirname(1) and basename(1)).
extname: extract the extension of the file name (without the '.').
barename: extract the base file name without the extension.
getnum[:TYPE]: extract a number from the field. TYPE is optional single letter option n/i/d/p/h/o (see examples below).
cut: copy input field to output field (similar to cut(1)).

Numeric Grouping operations

sum: sum the of values
min: minimum value
max: maximum value
absmin: minimum of the absolute values
absmax: maximum of the absolute values
range: the values range (max-min)

Textual/Numeric Grouping operations

count: count number of elements in the group
first: the first value of the group
last: the last value of the group
rand: one random value from the group
unique: comma-separated sorted list of unique values
collapse: comma-separated list of all input values
countunique: number of unique/distinct values

Statistical Grouping operations

A p/s prefix indicates the varient: population or sample. Typically, the sample variant is equivalent with GNU R's internal functions (e.g datamash's sstdev operation is equivalent to R's sd() function).

mean: mean of the values
geomean: geometric mean of the values
harmmean: harmonic mean of the values
trimmean[:PERCENT]: trimmed mean of the values. PERCENT should be between 0 and 0.5. (trimmean:0 is equivalent to mean. trimmean:0.5 is equivalent to median).
median: median value
q1: 1st quartile value
q3: 3rd quartile value
iqr: inter-quartile range
perc[:PERCENTILE]: percentile value PERCENTILE (defaults to 95).
mode: mode value (most common value)
antimode: anti-mode value (least common value)
pstdev/sstdev: population/sample standard deviation
pvar/svar: population/sample variance
mad: median absolute deviation, scaled by constant 1.4826 for normal distributions
madraw: median absolute deviation, unscaled
pskew/sskew: skewness of the group values x reported by 'sskew' and 'pskew' operations:

          x > 0       -  positively skewed / skewed right
      0 > x           -  negatively skewed / skewed left
          x > 1       -  highly skewed right
      1 > x >  0.5    -  moderately skewed right
    0.5 > x > -0.5    -  approximately symmetric
   -0.5 > x > -1      -  moderately skewed left
     -1 > x           -  highly skewed left

pkurt/skurt: excess Kurtosis of the group
jarque/dpo: p-value of the Jarque-Beta (jarque) and D'Agostino-Pearson Omnibus (dpo) tests for normality: null hypothesis is normality; low p-Values indicate non-normal data; high p-Values indicate null-hypothesis cannot be rejected.
pcov/scov [X:Y]: covariance of fields X and Y
ppearson/spearson [X:Y]: Pearson product-moment correlation coefficient [Pearson's R] of fields X and Y

EXAMPLES

Basic usage

Print the sum and the mean of values from field 1:

$ seq 10 | datamash sum 1 mean 1
55  5.5

Group input based on field 1, and sum values (per group) on field 2:

$ cat example.txt
A  10
A  5
B  9
B  11

$ datamash -g 1 sum 2 < example.txt
A  15
B  20

$ datamash groupby 1 sum 2 < example.txt
A  15
B  20

Unsorted input must be sorted (with '-s'):

$ cat example.txt
A  10
C  4
B  9
C  1
A  5
B  11

$ datamash -s -g1 sum 2 < example.txt
A  15
B  20
C  5

Which is equivalent to:

$ cat example.txt | sort -k1,1 | datamash -g 1 sum 2

Header lines

Use -h (--headers) if the input file has a header line:

# Given a file with student name, field, test score...
$ head -n5 scores_h.txt
Name           Major            Score
Shawn          Engineering      47
Caleb          Business         87
Christian      Business         88
Derek          Arts             60

# Calculate the mean and standard devian for each major
$ datamash --sort --headers --group 2 mean 3 pstdev 3 < scores_h.txt

 (or use short form)

$ datamash -sH -g2 mean 3 pstdev 3 < scores_h.txt

 (or use named fields)

$ datamash -sH -g Major mean Score pstdev Score < scores_h.txt
GroupBy(Major)    mean(Score)   pstdev(Score)
Arts              68.9          10.1
Business          87.3           4.9
Engineering       66.5          19.1
Health-Medicine   90.6           8.8
Life-Sciences     55.3          19.7
Social-Sciences   60.2          16.6

Field names must be escaped with a backslash if they start with a digit or contain special characters (dash/minus, colons, commas). Note the interplay between escaping with backslash and shell quoting. The following equivalent command sum the values of a field named "FOO-BAR":

$ datamash -H sum FOO\\-BAR < input.txt
$ datamash -H sum 'FOO\-BAR' < input.txt
$ datamash -H sum "FOO\\-BAR" < input.txt

Skipping comment lines

Use -C (--skip-comments) to skip lines starting with '#' or ';' characters (and optional whitespace before them):

$ cat in.txt
 #foo   3
bar     5
;baz    7

$ datamash sum 2 < in.txt
15

$ datamash -C sum 2 < in.txt
5

Multiple fields

Use comma or dash to specify multiple fields. The following are equivalent:

$ seq 9 | paste - - -
1   2   3
4   5   6
7   8   9

$ seq 9 | paste - - - | datamash sum 1 sum 2 sum 3
12  15  18

$ seq 9 | paste - - - | datamash sum 1,2,3
12  15  18

$ seq 9 | paste - - - | datamash sum 1-3
12  15  18

Rounding

The following demonstrate the different rounding operations:

$ ( echo X ; seq -1.25 0.25 1.25 ) \
      | datamash --full -H round 1 ceil 1 floor 1 trunc 1 frac 1

  X     round(X)  ceil(X)  floor(X)  trunc(X)   frac(X)
-1.25   -1        -1       -2        -1         -0.25
-1.00   -1        -1       -1        -1          0
-0.75   -1         0       -1         0         -0.75
-0.50   -1         0       -1         0         -0.5
-0.25    0         0       -1         0         -0.25
 0.00    0         0        0         0          0
 0.25    0         1        0         0          0.25
 0.50    1         1        0         0          0.5
 0.75    1         1        0         0          0.75
 1.00    1         1        1         1          0
 1.25    1         2        1         1          0.25

Reversing fields

$ seq 6 | paste - - | datamash reverse
2    1
4    3
6    5

Transposing a file

$ seq 6 | paste - - | datamash transpose
1    3    5
2    4    6

Removing Duplicated lines

Remove lines with duplicate key value from field 1 (Unlike first,last operations, rmdup is much faster and does not require sorting the file with -s):

# Given a list of files and sample IDs:
$ cat INPUT
SampleID  File
2         cc.txt
3         dd.txt
1         ab.txt
2         ee.txt
3         ff.txt

# Remove lines with duplicated Sample-ID (field 1):
$ datamash rmdup 1 < INPUT

# or use named field:
$ datamash -H rmdup SampleID < INPUT
SampleID  File
2         cc.txt
3         dd.txt
1         ab.txt

Checksums

Calculate the sha1 hash value of each TXT file, after calculating the sha1 value of each file's content:

$ sha1sum *.txt | datamash -Wf sha1 2

Check file structure

Check the structure of the input file: ensure all lines have the same number of fields, or expected number of lines/fields:

$ seq 10 | paste - - | datamash check && echo ok || echo fail
5 lines, 2 fields
ok

$ seq 13 | paste - - - | datamash check && echo ok || echo fail
line 4 (3 fields):
  10  11  12
line 5 (2 fields):
  13
datamash: check failed: line 5 has 2 fields (previous line had 3)
fail

$ seq 10 | paste - - | datamash check 2 fields 5 lines
5 lines, 2 fields

$ seq 10 | paste - - | datamash check 4 fields
line 1 (2 fields):
  1     2
datamash: check failed: line 1 has 2 fields (expecting 4)

$ seq 10 | paste - - | datamash check 7 lines
datamash: check failed: input had 5 lines (expecting 7)

Cross-Tabulation

Cross-tabulation compares the relationship between two fields. Given the following input file:

$ cat input.txt
a    x    3
a    y    7
b    x    21
a    x    40

Show cross-tabulation between the first field (a/b) and the second field (x/y) - counting how many times each pair appears (note: sorting is required):

$ datamash -s crosstab 1,2 < input.txt
     x    y
a    2    1
b    1    N/A

An optional grouping operation can be used instead of counting:


$ datamash -s crosstab 1,2 sum 3 < input.txt
     x    y
a    43   7
b    21   N/A

$ datamash -s crosstab 1,2 unique 3 < input.txt
     x    y
a    3,40 7
b    21   N/A

Binning numeric values

Bin input values into buckets of size 5:

$  ( echo X ; seq -10 2.5 10 ) \
      | datamash -H --full bin:5 1
    X  bin(X)
-10.0    -15
 -7.5    -10
 -5.0    -10
 -2.5     -5
  0.0      0
  2.5      0
  5.0      5
  7.5      5
 10.0     10

Binning string values

Hash any input value into a numeric integer. A typical usage would be to split an input file into N chunks, ensuring that all values of a certain key will be stored in the same chunk:

$ cat input.txt
PatientA   10
PatientB   11
PatientC   12
PatientA   14
PatientC   15


Each patient ID is hashed into a bin between 0 and 9
and printed in the last field:


$ datamash --full strbin 1 < input.txt
PatientA   10    5
PatientB   11    6
PatientC   12    7
PatientA   14    5
PatientC   15    7


Splitting the input into chunks can be done with awk:


$ cat input.txt \
    | datamash --full strbin 1 \
    | awk '{print > $NF ".txt"}'

Extracting numbers with getnum

The 'getnum' operation extracts a numeric value from the field:

$ echo zoom-123.45xyz | datamash getnum 1
123.45

getnum accepts an optional single-letter TYPE option:
    getnum:n - natural numbers (positive integers, including zero)
    getnum:i - integers
    getnum:d - decimal point numbers
    getnum:p - positive decimal point numbers (this is the default)
    getnum:h - hex numbers
    getnum:o - octal numbers

Examples:

$ echo zoom-123.45xyz | datamash getnum 1
123.45

$ echo zoom-123.45xyz | datamash getnum:n 1
123

$ echo zoom-123.45xyz | datamash getnum:i 1
-123

$ echo zoom-123.45xyz | datamash getnum:d 1
123.45

$ echo zoom-123.45xyz | datamash getnum:p 1
-123.45

# Hex 0x123 = 291 Decimal
$ echo zoom-123.45xyz | datamash getnum:h 1
291

# Octal 0123 = 83 Decimal
$ echo zoom-123.45xyz | datamash getnum:o 1
83