Skip to content(if available)orjump to list(if available)

XAN: A Modern CSV-Centric Data Manipulation Toolkit for the Terminal

XAN: A Modern CSV-Centric Data Manipulation Toolkit for the Terminal

24 comments

·March 27, 2025
Visit

dlkmp

Can't help but thinking how handy PowerShell is out of the box for tasks like this.

Translating the examples from the ReadMe, having read the file with:

  $medias = Get-Content .\medias.csv | ConvertFrom-Csv
Previewing the file in the terminal

  xan view medias.csv
  $medias | Format-Table
Reading a flattened representation of the first row

  xan flatten -c medias.csv
  $medias | Format-List
Searching for rows

  xan search -s outreach internationale medias.csv | xan view
  $medias | Where-Object { $_.outreach -eq "internationale" } | Format-Table
Selecting some columns

  xan select foundation_year,name medias.csv | xan view
  $medias | Select-Object -Property foundation_year, name | Format-Table
Sorting the file

  xan sort -s foundation_year medias.csv | xan view -s name,foundation_year
  $medias | Sort-Object -Property foundation_year | Select-Object -Property name, foundation_year | Format-Table
Deduplicating the file on some column

  # Some medias of our corpus have the same ids on mediacloud.org
  xan dedup -s mediacloud_ids medias.csv | xan count && xan count medias.csv
  $medias | Select-Object -ExpandProperty mediacloud_ids -Unique | Measure-Object; $medias | Measure-Object -Property mediacloud_ids
Computing frequency tables

  xan frequency -s edito medias.csv | xan view
  $medias | Group-Object -Property edito | Sort-Object -Property Count -Descending
It's probably orders of magnitude slower, and of course, plotting graphs and so on gets tricky. But for the simple type of analysis I typically do, it's fast enough, I don't need to learn an extra tool, and the auto-completion of column/property names is very convenient.

account-5

I find Nushell even better for these usecases:

    $medias = open .\medias.csv
The above is the initial read and format into table.

I'm currently on my phone so can't go through all the examples, but knowing both PS and nu, nu has the better syntax.

EDIT:

Get data and view in table:

    let $medias = http get https://github.com/medialab/corpora/raw/master/polarisation/medias.csv
    $medias
Get headers:

    $medias | columns
Get count of rows:

   $medias | length
Get flattened, slight more convoluted (caveat there might be a better way):

    $medias | each {print $in}
Search rows:

    $medias | where $it.outreach == 'internationale'
Select columns:

    $medias | select foundation_year name
Sort file:

    $medias | select foundation_year name | sort-by foundation_year
Dedup based on column:

    $medias | uniq-by mediacloud_ids
Computing frequency and histogram

    $medias | histogram edito

SwamyM

Yes, I find PowerShell is criminally underrated for these type of tasks. Even though it's open source and cross-platform, the stigma from it's Windows-centric days is hard to overcome.

pradeepchhetri

I prefer to use clickhouse-local for all my CSV needs as I don't need to learn a new language (or cli flags) and can just leverage SQL.

    clickhouse local --file medias.csv --query "SELECT edito, count() AS count from table group by all order by count FORMAT PrettyCompact"

   ┌─edito──────┬─count─┐
   │ agence     │     1 │
   │ agrégateur │    10 │
   │ plateforme │    14 │
   │ individu   │    30 │
   │ media      │   423 │
   └────────────┴───────┘
With clickhouse-local, I can do lot more as I can leverage full power of clickhouse.

SoftTalker

I used to use q for this sort of thing. Not sure if there are better choices now as it have been a few years.

https://harelba.github.io/q/

rixed

How does it compare with duckdb, which I usualy resort to? What I like with duckdb is that it's a single binary, no server needed, and it's been happy so far with all the CSV file I've thrown at it.

pradeepchhetri

clickhouse-local is similar to duckdb, you don't need a clickhouse-server running in order to use clickhouse-local. You just need to download the clickhouse binary and start using it.

  clickhouse local
  ClickHouse local version 25.4.1.1143 (official build).

  :)
There are few benefits of using clickhouse-local since ClickHouse can just do lot more than DuckDB. One such example is handling compressed files. ClickHouse can handle compressed files with formats ranging from zstd, lz4, snappy, gz, xz, bz2, zip, tar, 7zip.

  clickhouse local --query "SELECT count() FROM file('top-1m-2018-01-10.csv.zip :: *.csv')"
  1000000
Also clickhouse-local is much more efficient in handling big csv files[0]

[0]: https://www.vantage.sh/blog/clickhouse-local-vs-duckdb

sitkack

I use SQLite in a similar manner, but I'll have to check this out.

jgord

I was going to mention BurntSushis excellent xsv, also written in rust ... but I see xan mantions its a fork + rewrite of xsv.

doug_durham

I use Pandas for most of my CSV work. It's super fast and very powerful. There's a bit of a learning curve. I can then use Python scripts to manipulate massive CSV files.

gpvos

I tend to use csvkit for more complicated transformations, and OCaml's csvtool[0] for the simpler ones. For intermediate transformations I wrote my own csved[1] script, which reads for every line of a CSV reads it into @F, applies a Perl expression to that array, then writes it out. With the -h option you can also use the %F hash to access fields by name. It's very fast.

It looks like xsv and xan are in the "csvkit but faster" niche, which is nice, but now I must learn another set of commands.

And there are now many more recent utilities called csvtool, including a Perl and a Python one.

[0] https://github.com/Chris00/ocaml-csv

[1] https://github.com/gpvos/csved

teddyh

People will do anything to avoid learning SQL.

mike-the-mikado

I recently came across miller (https://github.com/johnkerl/miller). I don't know how these tools compare.

evolve2k

Something that would be insanely useful is if your tool could be users to do validations.

For example being able to define data types for each column and say required columns. And then run your tool as a validator and take the errors as an array that’d be amazing!

Coming from a dev who’s just over processing CSV files back into my apps.

hantusk

Reading CSV into a duckdb table will give you that, along with a table for the errors and reason for error: https://duckdb.org/docs/stable/data/csv/reading_faulty_csv_f...

Could definitely be done as a small little bash script

ranger_danger

https://digital-preservation.github.io/csv-validator/

crashabr

You should take a look at the Frictionless data library https://framework.frictionlessdata.io/docs/guides/validating...

sevg

This looks really nice to use!

There are a lot of tools that one can use on Linux cli to work with csv. But many of them have become unmaintained. Or have terrible docs. Or have really awkward usage (looking at you, “yq”).

ctippett

See also: csvkit (https://csvkit.readthedocs.io)

strunz

Yep, been using csvkit for years and it's so great. There's nothing I haven't been able to do, is there some reason to use Xan instead?

aorth

I use both csvkit and xsv. The syntax for csvkit is a bit easier for most of my uses, but xsv is way faster when I have larger files.

I know someone who uses csvtk (Golang), but haven't tried it yet. https://github.com/shenwei356/csvtk

reagle

This could be a nice pre-filter for VisiData…?

andrea76

I am using Goawk with good results