PIP in LibreOffice 2025.8 May break many things including numpy and Pandas

English
, ,
#1

I am the author of LibrePythonista. Today I was testing the extension in LibreOffice Developer.

Version: 25.8.0.0.alpha1 (X86_64) / LibreOffice Community
Build ID: a2896e060b54ce6ce36e3e29218fac40d2919c74
CPU threads: 4; OS: Windows 10 X86_64 (build 19045); UI render: Skia/Raster; VCL: win
Locale: en-US (en_US); UI: en-US
Calc: CL threaded

This version of LibreOffice contains a pip installer. However, the pip installer is not a complete version.
It is missing several files such as the following in the C:\Program Files\LibreOfficeDev 25\program\python-core-3.11.12\lib\pip\_vendor\distlib folder:

  • t32.exe
  • t64.exe
  • t64-arm.exe
  • w32.exe
  • w64.exe
  • w64-arm.exe

When I try t install Pandas Manually in PowerShell I get the following Error:

& "C:\Program Files\LibreOfficeDev 25\program\python.exe" -m pip install pandas

...
Installing collected packages: pytz, tzdata, six, numpy, python-dateutil, pandas
...
  File "C:\Program Files\LibreOfficeDev 25\program\python-core-3.11.12\lib\pip\_vendor\distlib\scripts.py", line 339, in _make_script
    self._write_script(scriptnames, shebang, script, filenames, ext)
  File "C:\Program Files\LibreOfficeDev 25\program\python-core-3.11.12\lib\pip\_vendor\distlib\scripts.py", line 260, in _write_script
    launcher = self._get_launcher('t')
               ^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Program Files\LibreOfficeDev 25\program\python-core-3.11.12\lib\pip\_vendor\distlib\scripts.py", line 419, in _get_launcher
    raise ValueError(msg)
ValueError: Unable to find resource t64.exe in package pip._vendor.distlib

The t64.exe file and maybe more seem to be required to install Pandas.

I suspect any existing methods to install Pandas and some other packages into LibreOffice will fail when embedded python with pip is used.

This also affect LibrePythonista is it also installs Numpy, Pandas and more, or at least did on previous versions.
With the built in pip LIbrePythonista can no longer Install the required packages. It is not really possible or recommended to install a second pip to get around this.

I tired a workaround just to see if it could work and it does.
My workaround was to download embeddable python 3.11.12 and extract is contents.
Then I replaced C:\Program Files\LibreOfficeDev 25\program\python-core-3.11.12\lib\pip with the pip from the embedded version.

I tested installing LibrePythonista and Pandas again and it worked.

Any thoughts on this?

#2

For those who want to see it I had a conversation with Gemini about this issue.

How to make simple divison problem using random numbers
#3

Your link points wrongly to a 3.13.3 -version!

Drop your selfmade ransom-ware ( especially the pandas- , numpy-extension ) and help the devs with debugging the windows-version of the next release (LO25.8)

#4

How does it meet that definition?

#5

It’s an analogy. With the extension being installed, every loaded spreadsheet gets extension data wrapped into the document structure, regardless if the document uses the extension or not. Apart from that, it is very difficult to get any practical use from this extension. The software manipulates a whole class of files without any obvious benefit to the user.

1 Like
#6

In addition to what @Villeroy already wrote, the so-called numpy and pandas extensions primarily install a lot of packages that are definitely NOT needed by either numpy or pandas, especially for python beginners this is all difficult to understand and difficult to uninstall completely.

Conclusion: Use pip to install “third-party python packages” or, in case of doubt, APSO in Libreoffice.

#7

Bloatware is probably a more appropriate term.
.
My concern with numpy and pandas is passing financial data through them screws up data accuracy and adds all sorts of artefacts.

#8

Your link points wrongly to a 3.13.3 -version!


In My testing it was the correct version

Drop your selfmade ransom-ware ( especially the pandas- , numpy-extension ) and help the devs with debugging the windows-version of the next release (LO25.8)


Ransomware is a type of malware that encrypts the victim’s personal data until a ransom is paid. Difficult-to-trace digital currencies such as paysafecard or Bitcoin and other cryptocurrencies are commonly used for the ransoms, making tracing and prosecuting the perpetrators difficult. Sometimes the original files can be retrieved without paying the ransom due to implementation mistakes, leaked cryptographic keys or a complete lack of encryption in the ransomware.


To call my software or any other python package in this context is a complete misrepresentation and not even close to correct. Also note that all my software extensions and related source code is completely open source. Anyone can audit my software and I encourage them to do so.


You comment makes the assumption that I am not helping the community. Many, many thousands who use the various extensions I have created would likely disagree. To suggest that I do not help the community is a flat-out lie and insulting.
I have spent that last five years and many thousands of hours working on implementing python into LibreOffice in various ways and to date I have received the grand total of $0 dollars. Last time I checked installing an extension is completely optional. A far cry I am sure of someone who would be so malicious as to create ransomware.


@karolus I don’t know you. I have never attracted you in any way that I am aware of. This is another one of your attacks on my work and my character and this behavior is simply not acceptable. I will pursue available measure to stop your attack on my good works.


Apart from all of that this is not even about my extensions. I am sure I made it clear.
I am stating that adding Pip to LibreOffice in its current form will negate the purpose of including pip with LibreOffice in the first place.
Any package including pandas will not install into LibreOffice when it has missing pip files. I have outlined all of this. Form more clarity read the conversation with Gemini.


In my estimation this is a huge flaw and will break many things for many users, not just users of my extension. I can write workarounds for my extension but this is still a huge issue.


In addition to what @Villeroy already wrote, the so-called numpy and pandas extensions primarily install a lot of packages that are definitely NOT needed by either numpy or pandas, especially for python beginners this is all difficult to understand and difficult to uninstall completely.


Where is your evidence for this? The Numpy and Pandas extension that I wrote only install what is needed and also have uninstallers.


Please lets address the actual issues this post was started for. Installing Numpy or Pandas will also not be possible with APSO due to the issue I have outlined here. Including Pip in its current form will break installing many packages for many installers not just the ones I have created.

#9

@vib I agree that calling your software “ransomware” is unacceptable. People may have their opinions on the usefulness of other people’ products, or on some architectural decisions (I myself am a bit annoyed by the behavior of installer of your extensions, which I needed to test several times when tried to understand some problems that you reported - which may be because we don’t provide better means for you to use); but that is different from such attacks.

I am not sure it’s the proper place to address it here. The change was in tdf#162786; and as not a newbie here, you should realize that such problems as you mention should be addressed also in the bug tracker.

1 Like
#10

Who moderates the moderators on here?

#11

Thanks @mikekaganski I submitted a new bug report: tdf#166746


tdf#162786 is already resolved.


I myself am a bit annoyed by the behavior of installer of your extensions, which I needed to test several times when tried to understand some problems that you reported - which may be because we don’t provide better means for you to use

Any feedback or suggestion you have are very welcomed.

2 Likes

Impressum (Legal Info) | Datenschutzerklärung (Privacy Policy)   Content on this site is licensed under a Creative Commons Attribution Share Alike 3.0 license.