pipx behave different than pip if package already exists

2021-09-23T00:00:00+02:00

I decided to put as much as possible of my local config into an Ansible Playbook. Makes it easier to track what I install and change, also to setup from scratch should it necessary.

In the past I mainly used pip in combination with venv to install Python stuff. Changing for most of it to pipx seems logical, it will take care about the necessary venvs on it's own.

Just had to learn that the return code of pip is 0 but from pipx it's 1 in case a package already exists.

Why does it matter? Not the most elegant solution but I decided to just use command to trigger pip / pipx from my Playbook:

- name: Python Package Installation (pip)
  command: "python3 -m pip install --user {{ item }}"
  changed_when: false
  loop:
    - pipx
    - nox

- name: Python Package Installation (pipx)
  command: "pipx install {{ item }}"
  changed_when: false
  loop:
    - poetry
    - duplicity
    - pre-commit

No problem with pip, but pipx failed after the first run.

There was an older GitHub Issue, based on the comments, pipx was supposed to behave smilar as pip and also use return code 0 since v0.13 already. After some research I found the Pull Request that reverted the earlier change during some re-factoring.

Thanks to the great improvements in that PR, it was just a very small change necessary to restore the functionality introduced in Commit 11b853e:

diff --git a/docs/changelog.md b/docs/changelog.md
index 0a95ce15..67d14929 100644
--- a/docs/changelog.md
+++ b/docs/changelog.md
@@ -1,6 +1,7 @@
 dev

 - Fixed `pipx list` output phrasing to convey that python version displayed is the one with which package was installed.
+- Fixed `pipx install` to provide return code 0 if venv already exists, similar to pip’s behavior. (#736)

 0.16.4

diff --git a/src/pipx/constants.py b/src/pipx/constants.py
index 4fe2d58b..11fc013f 100644
--- a/src/pipx/constants.py
+++ b/src/pipx/constants.py
@@ -21,7 +21,7 @@
 # pipx shell exit codes
 EXIT_CODE_OK = ExitCode(0)
 EXIT_CODE_INJECT_ERROR = ExitCode(1)
-EXIT_CODE_INSTALL_VENV_EXISTS = ExitCode(1)
+EXIT_CODE_INSTALL_VENV_EXISTS = ExitCode(0)
 EXIT_CODE_LIST_PROBLEM = ExitCode(1)
 EXIT_CODE_UNINSTALL_VENV_NONEXISTENT = ExitCode(1)
 EXIT_CODE_UNINSTALL_ERROR = ExitCode(1)
diff --git a/tests/test_install.py b/tests/test_install.py
index 941f4c01..0eb2d9cc 100644
--- a/tests/test_install.py
+++ b/tests/test_install.py
@@ -109,7 +109,7 @@ def test_install_no_packages_found(pipx_temp_env, capsys):

 def test_install_same_package_twice_no_force(pipx_temp_env, capsys):
     assert not run_pipx_cli(["install", "pycowsay"])
-    assert run_pipx_cli(["install", "pycowsay"])
+    assert not run_pipx_cli(["install", "pycowsay"])
     captured = capsys.readouterr()
     assert (
         "'pycowsay' already seems to be installed. Not modifying existing installation"

Related Pull Request: https://github.com/pypa/pipx/pull/736

Merged into pipx/main on 25th September 2021

That's a good example why I love and prefer Open Source, I could fix the Problem on my own and share the improvement with the community, within a few days it was already merged and will be part of the next Release.

Create BMC Control-M usage reports from CLI instead GUI

2021-08-21T00:00:00+02:00

I had to solve a quite interesting problem recently, the Task was to generate Control-M usage reports on a monthly basis and send them via E-Mail.

Control-M provides a Solution for that: BMC Control-M Usage Reporting Tool

But it's a Java GUI Application that can't be automated or run in Batch mode. So I decided to write a small "wrapper" in Java which re-use the original Classes from the Control-M Usage Reporting Tool (Version 9.0.00).

That way the results are 100% identical without any (unsupported) customizing but it can be executed from the command line.

I want to make clear: All Control-M Copyrights belong to BMC, including the Java Code and Classes. All what my little wrapper does, is loading the Class which comes with the original Tool and call the same methods as it would be done in the GUI to generate the report.

Therefore I want to share the Code of my wrapper and the necessary cmd file (yes, all this is only running on Windows due to hardcoded backslashes in file paths in the used Class) to start it.

Downloading the required Jar file from BMC, Building the Project (I used "Export: Runnable JAR file" in Eclipse), adding the necessary data/env.txt and data/config.ini is up to you. I suggest you take a look into the directory structure and files of the BMC Control-M Usage Reporting Tool, it isn't that hard figure it out and get it working :)

ctmusagereport

package ctmusagereport;
import java.io.File;
import com.bmc.ctmem.usagetool.Enhancement.ZipFiles;
import com.bmc.ctmem.usagetool.dbaccess.AddOnsReport;
import com.bmc.ctmem.usagetool.dbaccess.DBAccess;
import com.bmc.ctmem.usagetool.dbaccess.DBAccessManager;
import com.bmc.ctmem.usagetool.dbaccess.EndpointReport;
import com.bmc.ctmem.usagetool.dbaccess.JobCountReport;
import com.bmc.ctmem.usagetool.dbaccess.Params;
public class Main {
    public static void main(String[] args) {
        // Used to Load Env and Config File also to Generate and Get Report Directory
        Params par = Params.instance();
        try {
            par.LoadEMEnvironmentsFromFile();
        } catch (Exception e) {
            e.printStackTrace();
        }
        try {
            par.LoadConfigParamFromFile();
        } catch (Exception e) {
            e.printStackTrace();
        }

        // Initiates DB Connection based on values in data/env.txt
        DBAccessManager dbam = DBAccessManager.instance();
        dbam.RefreshDBAccessManager();

        // Get DB Connection for Environment 0 (first line in env file)
        DBAccess con = dbam.GetDBAccess(0);
        con.GetEMVersionFromDB();

        // Creates new Directory in reports/ with datetime now ("yyyy-MM-dd-HHmmss")
        par.CreateReportDir();

        // Trigger Generation of all Reports identical to GUI (Wizard / FlowManager)
        new JobCountReport(par.GetReportDirName());
        new AddOnsReport(par.GetReportDirName());
        new EndpointReport(par.GetReportDirName(), true);

        // For easier handling, add all generated files to "reports.zip" in Generated Report Directory
        File[] fileReportsDir = new File(par.GetReportDirName()).listFiles();
        ZipFiles.Zip(fileReportsDir, par.GetReportDirName());
    }
}

ctmusagereport.cmd

java -cp "./*;ctmusagereport_lib/*" ctmusagereport.Main

I tested it successfully with "OpenJDK11U-jdk_x64_windows_hotspot_11.0.12_7" from https://adoptium.net/ on Windows Server 2016.

What I like about working with Control-M is that you can Script and Customize almost everything somehow, even worse that I had to find a workaround like that for the usage reporting, would be great if BMC would offer such an CLI option with the original Tool.

The Wombelix Post - Code

pipx behave different than pip if package already exists

Create BMC Control-M usage reports from CLI instead GUI