csvtk - a cross-platform, efficient and practical CSV/TSV toolkit


Similar to FASTA/Q format in field of Bioinformatics, CSV/TSV formats are basic and ubiquitous file formats in both Bioinformatics and data sicence.

People usually use spreadsheet softwares like MS Excel to do process table data. However it's all by clicking and typing, which is not automatically and time-consuming to repeat, especially when we want to apply similar operations with different datasets or purposes.

You can also accomplish some CSV/TSV manipulations using shell commands, but more codes are needed to handle the header line. Shell commands do not support selecting columns with column names either.

csvtk is convenient for rapid data investigation and also easy to be integrated into analysis pipelines. It could save you much time of writing Python/R scripts.

Table of Contents



35 subcommands in total.


Format conversion

Set operations






Download Page

csvtk is implemented in Go programming language, executable binary files for most popular operating systems are freely available in release page.

Method 1: Download binaries (latest stable/dev version)

Just download compressed executable file of your operating system, and decompress it with tar -zxvf *.tar.gz command or other tools. And then:

  1. For Linux-like systems

    1. If you have root privilege simply copy it to /usr/local/bin:

      sudo cp csvtk /usr/local/bin/
    2. Or copy to anywhere in the environment variable PATH:

      mkdir -p $HOME/bin/; cp csvtk $HOME/bin/
  2. For windows, just copy csvtk.exe to C:\WINDOWS\system32.

Method 2: Install via conda (latest stable version) Install-with-conda Anaconda Cloud downloads

conda install -c bioconda csvtk

Method 3: For Go developer (latest stable/dev version)

go get -u github.com/shenwei356/csvtk/csvtk

Method 4: For ArchLinux AUR users (may be not the latest)

yaourt -S csvtk


Note: The current version supports Bash only. This should work for *nix systems with Bash installed.


  1. run: csvtk genautocomplete

  2. create and edit ~/.bash_completion file if you don't have it.

    nano ~/.bash_completion

    add the following:

    for bcfile in ~/.bash_completion.d/* ; do
      . $bcfile

Compared to csvkit


Features csvtk csvkit Note
Read Gzip Yes Yes read gzip files
Fields ranges Yes Yes e.g. -f 1-4,6
Unselect fileds Yes -- e.g. -1 for excluding first column
Fuzzy fields Yes -- e.g. ab* for columns with name prefix "ab"
Reorder fields Yes Yes it means -f 1,2 is different from -f 2,1
Rename columns Yes -- rename with new name(s) or from existed names
Sort by multiple keys Yes Yes bash sort like operations
Sort by number Yes -- e.g. -k 1:n
Multiple sort Yes -- e.g. -k 2:r -k 1:nr
Pretty output Yes Yes convert CSV to readable aligned table
Unique data Yes -- unique data of selected fields
frequency Yes -- frequencies of selected fields
Sampling Yes -- sampling by proportion
Mutate fields Yes -- create new columns from selected fields
Repalce Yes -- replace data of selected fields

Similar tools:


More examples and tutorial.


  1. The CSV parser requires all the lines have same number of fields/columns. Even lines with spaces will cause error.
  2. By default, csvtk thinks your files have header row, if not, switch flag -H on.
  3. Column names better be unique.
  4. By default, lines starting with # will be ignored, if the header row starts with #, please assign flag -C another rare symbol, e.g. '$'.
  5. By default, csvtk handles CSV files, use flag -t for tab-delimited files.
  6. If " exists in tab-delimited files, use flag -l.
  7. Do not mix use digital fields and column names.


  1. Pretty result

    $ csvtk pretty names.csv
    id   first_name   last_name   username
    11   Rob          Pike        rob
    2    Ken          Thompson    ken
    4    Robert       Griesemer   gri
    1    Robert       Thompson    abc
    NA   Robert       Abel        123
  2. Summary of selected digital fields: num, sum, min, max, mean, stdev (stat2)

    $ cat digitals.tsv
    4       5       6
    1       2       3
    7       8       0
    8       1,000   4
    $ cat digitals.tsv | csvtk stat2 -t -H -f 1-3
    field   num     sum   min     max     mean    stdev
    1         4      20     1       8        5     3.16
    2         4   1,015     2   1,000   253.75   497.51
    3         4      13     0       6     3.25      2.5
  3. Select fields/columns (cut)

    • By index: csvtk cut -f 1,2
    • By names: csvtk cut -f first_name,username
    • Unselect: csvtk cut -f -1,-2 or csvtk cut -f -first_name
    • Fuzzy fields: csvtk cut -F -f "*_name,username"
    • Field ranges: csvtk cut -f 2-4 for column 2,3,4 or csvtk cut -f -3--1 for discarding column 1,2,3
    • All fields: csvtk cut -F -f "*"
  4. Search by selected fields (grep) (matched parts will be highlighted as red)

    • By exactly matching: csvtk grep -f first_name -p Robert -p Rob
    • By regular expression: csvtk grep -f first_name -r -p Rob
    • By pattern list: csvtk grep -f first_name -P name_list.txt
    • Remore rows containing missing data (NA): csvtk grep -F -f "*" -r -p "^$" -v
  5. Rename column names (rename and rename2)

    • Setting new names: csvtk rename -f A,B -n a,b or csvtk rename -f 1-3 -n a,b,c
    • Replacing with original names by regular express: cat ../testdata/c.csv | ./csvtk rename2 -F -f "*" -p "(.*)" -r 'prefix_$1' for adding prefix to all column names.
  6. Edit data with regular expression (replace)

    • Remove Chinese charactors: csvtk replace -F -f "*_name" -p "\p{Han}+" -r ""
  7. Create new column from selected fields by regular expression (mutate)

    • In default, copy a column: csvtk mutate -f id
    • Extract prefix of data as group name (get "A" from "A.1" as group name): csvtk mutate -f sample -n group -p "^(.+?)\."
  8. Sort by multiple keys (sort)

    • By single column : csvtk sort -k 1 or csvtk sort -k last_name
    • By multiple columns: csvtk sort -k 1,2 or csvtk sort -k 1 -k 2 or csvtk sort -k last_name,age
    • Sort by number: csvtk sort -k 1:n or csvtk sort -k 1:nr for reverse number
    • Complex sort: csvtk sort -k region -k age:n -k id:nr
  9. Join multiple files by keys (join)

    • All files have same key column: csvtk join -f id file1.csv file2.csv
    • Files have different key columns: csvtk join -f "username;username;name" names.csv phone.csv adress.csv -k
  10. Filter by numbers (filter)

    • Single field: csvtk filter -f "id>0"
    • Multiple fields: csvtk filter -f "1-3>0"
    • Using --any to print record if any of the field satisfy the condition: csvtk filter -f "1-3>0" --any
    • fuzzy fields: csvtk filter -F -f "A*!=0"
  11. Filter rows by awk-like arithmetic/string expressions (filter2)

    • Using field index: csvtk filter2 -f '$3>0'
    • Using column names: csvtk filter2 -f '$id > 0'
    • Both arithmetic and string expressions: csvtk filter2 -f '$id > 3 || $username=="ken"'
    • More complicated: csvtk filter2 -H -t -f '$1 > 2 && $2 % 2 == 0'
  12. Ploting

    • plot histogram with data of the second column: csvtk -t plot hist testdata/grouped_data.tsv.gz -f 2 | display histogram.png
    • plot boxplot with data of the "GC Content" (third) column, group information is the "Group" column. csvtk -t plot box testdata/grouped_data.tsv.gz -g "Group" -f "GC Content" --width 3 | display boxplot.png
    • plot horiz boxplot with data of the "Length" (second) column, group information is the "Group" column. csvtk -t plot box testdata/grouped_data.tsv.gz -g "Group" -f "Length" --height 3 --width 5 --horiz --title "Horiz box plot" | display boxplot2.png
    • plot line plot with X-Y data csvtk -t plot line testdata/xy.tsv -x X -y Y -g Group | display lineplot.png
    • plot scatter plot with X-Y data csvtk -t plot line testdata/xy.tsv -x X -y Y -g Group --scatter | display scatter.png


We are grateful to Zhiluo Deng and Li Peng for suggesting features and reporting bugs.

Thanks Albert Vilella for features suggestion, which makes csvtk feature-rich。


create an issue to report bugs, propose new functions or ask for help.

Or leave a comment.


MIT License