313
votes

I'm looking for a way to convert xlsx files to csv files on Linux.

I do not want to use PHP/Perl or anything like that since I'm looking at processing several millions of lines, so I need something quick. I found a program on the Ubuntu repos called xls2csv but it will only convert xls (Office 2003) files (which I'm currently using) but I need support for the newer Excel files.

Any ideas?

10
Thinking that anything implemented with a scripting language is going to be slow by nature seems... a little misguided, particularly since the interesting libraries in those languages tend to have backends written in C.Charles Duffy
Excel used to be limited to 65536 rows. Now it's 1,048,576 (support.microsoft.com/kb/120596). it's going to be tough to fit "sever millions of lines" in it. just saying...Pavel Veller
@Pavel could be over several files.Charles Duffy
...personally, I'd do this using the xlsv library for Python, but since scripting-based approaches are described as out of the question... shrug. (How is it a programming question if programmatic tools are excluded from the answer?)Charles Duffy
@CharlesDuffy I'm currently using a PHP library to do this, and what takes xls2csv 1 second to do, takes php 10 minutes to do. Literally.user1390150

10 Answers

273
votes

The Gnumeric spreadsheet application comes with a command line utility called ssconvert that can convert between a variety of spreadsheet formats:

$ ssconvert Book1.xlsx newfile.csv
Using exporter Gnumeric_stf:stf_csv

$ cat newfile.csv 
Foo,Bar,Baz
1,2,3
123.6,7.89,
2012/05/14,,
The,last,Line

To install on Ubuntu:

apt-get install gnumeric

To install on Mac:

brew install gnumeric
150
votes

You can do this with LibreOffice:

libreoffice --headless --convert-to csv $filename --outdir $outdir

For reasons not clear to me, you might need to run this with sudo. You can make LibreOffice work with sudo without requiring a password by adding this line to you sudoers file:

users ALL=(ALL) NOPASSWD: libreoffice
150
votes

If you already have a Desktop environment then I'm sure Gnumeric / LibreOffice would work well, but on a headless server (such as Amazon Web Services), they require dozens of dependencies that you also need to install.

I found this Python alternative:

https://github.com/dilshod/xlsx2csv

$ easy_install xlsx2csv
$ xlsx2csv file.xlsx > newfile.csv

Took 2 seconds to install and works like a charm.

If you have multiple sheets you can export all at once, or one at a time:

$ xlsx2csv file.xlsx --all > all.csv
$ xlsx2csv file.xlsx --all -p '' > all-no-delimiter.csv
$ xlsx2csv file.xlsx -s 1 > sheet1.csv

He also links to several alternatives built in Bash, Python, Ruby, and Java.

40
votes

In bash, I used this libreoffice command to convert all my xlsx files in the current directory:

for i  in *.xlsx; do  libreoffice --headless --convert-to csv "$i" ; done

Close all your Libre Office open instances before executing, or it will fail silently.

The command takes care of spaces in the filename.

Tried again some years later, and it didn't work. This thread gives some tips, but the quickest solution was to run as root (or running a sudo libreoffice). Not elegant, but quick.

Use the command scalc.exe in Windows

36
votes

Use csvkit

in2csv data.xlsx > data.csv

For details check their excellent docs

11
votes

Another option would be to use R via a small bash wrapper for convenience:

xlsx2txt(){
echo '
require(xlsx)
write.table(read.xlsx2(commandArgs(TRUE)[1], 1), stdout(), quote=F, row.names=FALSE, col.names=T, sep="\t")
' | Rscript --vanilla - $1 2>/dev/null
}

xlsx2txt file.xlsx > file.txt
8
votes

If .xlsx file has many sheets, -s flag can be used to get the sheet you want. For example:

xlsx2csv "my_file.xlsx" -s 2 second_sheet.csv

second_sheet.csv would contain data of 2nd sheet in my_file.xlsx.

6
votes

Using the Gnumeric spreadsheet application which comes which a commandline utility called ssconvert is indeed super simple:

find . -name '*.xlsx' -exec ssconvert -T Gnumeric_stf:stf_csv {} \;

and you're done!

4
votes

If you are OK to run Java command line then you can do it with Apache POI HSSF's Excel Extractor. It has a main method that says to be the command line extractor. This one seems to just dump everything out. They point out to this example that converts to CSV. You would have to compile it before you can run it but it too has a main method so you should not have to do much coding per se to make it work.

Another option that might fly but will require some work on the other end is to make your Excel files come to you as Excel XML Data or XML Spreadsheet of whatever MS calls that format these days. It will open a whole new world of opportunities for you to slice and dice it the way you want.

2
votes

As others said, libreoffice can convert xls files to csv. The problem for me was the sheet selection.

This libreoffice Python script does a fine job at converting a single sheet to CSV.

Usage is:

./libreconverter.py File.xls:"Sheet Name" output.csv

The only downside (on my end) is that --headless doesn't seem to work. I have a LO window that shows up for a second and then quits.
That's OK with me, it's the only tool that does the job rapidly.