I'm trying to install yi.
I've got the following stack.yaml file:
resolver: lts-8.12
extra-deps:
- yi-frontend-pango-0.13.5
A stack install yi gives me an error of: Stack looks for packages in the directories configured in the 'packages' variable defined in your stack.yaml.
Am I only able to install this via editing my global stack.yaml config?
The full error should be something like
Stack looks for packages in the directories configured in the
'packages' variable defined in your stack.yaml The current entry
points to /home/mgsloan/fpco/test-stack/stack-overflow-43575553/ but
no .cabal file could be found there.
The issue is that the packages field defaults to the current directory, since it's a common use case to use it with just one package. You can override this by doing the following:
packages: []
Related
I am trying to Provision VMs on KVM with Terraform.
one of the steps in installations is to download and install the provider buy the command:
go install github.com/dmacvicar/terraform-provider-libvirt#latest
but it errors:
The go.mod file for the module providing named packages contains one or
more replace directives. It must not contain directives that would cause it to be interpreted differently than if it were the main module.
I didn't find a solution, is someone has faced it?
thank you
As JimB notice in comments:
If there are replace or exclude directives in the module, the correct
installation method is to clone the source and install it,
git clone github.com/dmacvicar/terraform-provider-libvirt
cd terraform-provider-libvirt
go install
Suppose I download a package using go get <Import path of package>. Now after x number of days that package has been updated remotely and now if I run again go get <Import path of package> will it be updated on my local too?
The go get <Import path of package> command will not update the local copy if there is already a local copy installed.
The go get documentation says:
The -u flag instructs get to use the network to update the named packages and their dependencies. By default, get uses the network to check out missing packages but does not use it to look for updates to existing packages.
Run go get -u <Import path of package> to get or update the local copy.
No. According to the go get documentation it will not. If you want to update a local package you have to use the -u flag.
$ go help get
......
The -u flag instructs get to use the network to update the named packages
and their dependencies. By default, get uses the network to check out
missing packages but does not use it to look for updates to existing packages.
Things change a bit if you use Go modules. Then the go.mod file is consulted, but by default it won't fetch the latest version automatically unless you tell it too. For this new behaviour please take a deeper look at the official docs https://github.com/golang/go/wiki/Modules#daily-workflow.
go version go1.11.4 darwin/amd64
GOPATH has been unset but was previously set to $HOME/Development/go
Project path is under $HOME/code/
I’m able to successfully (no errors at least) install the go-sql-driver/mysql package using the command
go get github.com/go-sql-driver/mysql#v1
When I include a reference to that package in an import statement
import(
_ "github.com/go-sql-driver/mysql")
in my code I see the error
could not import github.com/go-sql-driver/mysql (can’t find import:
“github.com/go-sql-driver/mysql”)
I have run go mod init in my project root and it creates a go.mod file. When I run the go get command I see a require statement added to that file for the package. But it seems the files for the package get installed in the default $HOME/go directory (since I've unset GOPATH).
Should I be doing things differently so that my import statement can find the newly installed package? Using modules shouldn't all the packages be installed in the project path somewhere?
Should I be doing things differently so that my import statement can find the newly installed package?
No. With modules there is no need to install the packages upfront at all.
Using modules shouldn't all the packages be installed in the project path somewhere?
No. They get downloaded somewhere in some format and used from that location but they are not "installed" like in the old GOPATH variant of go get.
Show output of go env and what go mod vendor produces.
I'm pretty sure I was doing things wrong. I was able to resolve this after referencing the following closely the steps documented at golang modules wiki. The summary is that there is no need to "install" a package via 'go get'. Instead simply make sure your project is initialized to use modules using the 'go mod init' command and then include the package name in an import statement. The next build event will pull down the package and all its dependencies.
When I try to use the Ansible's Composer module and paste the following task into my playbook.yml file I get an error.
playbook.yml
- name: Composer Install Site Dependencies
composer: command=install working_dir=/var/www/html
Error:
ERROR: composer is not a legal parameter in an Ansible task or handler
Ansible failed to complete successfully. Any error output should be
visible above. Please fix these errors and try again.
After some investigation I ran "anisble-doc --list" on the command line to see the available modules and "composer" is not listed. I am running Ansible version 1.5.4, do I have to add it separately?
As #user272735 indicated in the comments, this is an unreleased module- it's slated for the 1.6 release, which is under "active development". (admittedly it was originally slated for 1.4) You have a couple of options:
install ansible from the bleeding edge. See "running from source". (obviously, this is scary)
ninja-patch the file into your locally installed tree. (obviously, this is scary)
add the file into your local Ansible repo.
As "developing modules" says, a fourth option is to specify your library path via ANSIBLE_LIBRARY or --module-path. HOWEVER, this overrides your global library/module path. That isn't what you want to do unless you are providing every module.
adding into your repo
I'm assuming your repo is named "ansible" and is set up properly, like this:
ansible/
ansible/roles/
ansible/group_vars/
In that case, simply add a library directory at the top (the 'best practices' discusses this but not in the expected section):
ansible/
ansible/roles/
ansible/group_vars/
ansible/library/
Inside there, add the composer file in there. That makes its path/file the following:
ansible/library/composer
Note it is not composer.py or anything else. Also, it doesn't seem to need the +x bit, so no fussy worries there.
Once you do that, you can run Ansible commands as you'd expect. The composer module will simply be there.
I'm working on a support library for a large Python project which heavily uses relative imports by appending various project directories to sys.path.
Using The Hitchhiker's Guide to Packaging as a template I attempted to create a package structure which will allow me to do a local install, but can easily be changed to a global install later if desired.
One of the dependencies of my package is the pyasn1 package for the encoding and decoding of ASN.1 annotated objects. I have to include the pyasn1 library separately as the version supported by the CentOS 6.3 default repositories is one major version back and has known bugs that will break my custom package.
The top-level of the library structure is as follows:
MyLibrary/
setup.py
setup.cfg
LICENSE.txt
README.txt
MyCustomPackage/
pyasn1-0.1.6/
In my setup configuration file I define the install directory for my library to be a local directory called .lib. This is desirable as it allows me to do absolute imports by running the command import site; site.addsitedir("MyLibrary/.lib") in the project's main application without requiring our engineers to pass command line arguments to the setup script.
setup.cfg
[install]
install-lib=.lib
setup.py
setup(
name='MyLibrary',
version='0.1a',
package_dir = {'pyasn1': 'pyasn1-0.1.6/pyasn1'},
packages=[
'MyCustomPackage',
'pyasn1',
'pyasn1.codec',
'pyasn1.compat','
pyasn1.codec.ber',
'pyasn1.codec.cer',
'pyasn1.codec.der',
'pyasn1.type'
],
license='',
long_description=open('README.txt').read(),
data_files = []
)
The problem I've run into with doing the installation this way is that when my package tries to import pyasn1 it imports the global version and ignores the locally installed version.
As a possible workaround I have tried installing the pyasn1 package under a different name than the global package (eg pyasn1_0_1_6) by doing package_dir = {'pyasn1_0_1_6':'pyasn1-0.1.6/pyasn1'}. However, this fails since the imports used internally to the pyasn1 package do not use the pyasn1_0_1_6 name.
Is there some way to either a) force Python to import a locally installed package over a globally installed one or b) force a package to install under a different name?
Use virtualenv to ensure that your application runs in a fully known configuration which is independent from the OS version of libraries.
EDIT: a quick (unix) solution is setting the PYTHONPATH environment variable, which works just like PATH for Python modules (module loaded from first path in which is found, so simply append you directory at the beginning of the PYTHONPATH). Anwyay, I strongly recommend you to proceed with virtualenv, since it was specifically engineered for handling situations like the one you are facing.
Rationale
The process is easily automatable if you write a setuptools script specifying dependencies with install_requires. For a complete example, refer to this one I wrote
Setup
Note that you can easily insert the steps below in a setup.sh shell script.
First create a virtualenv and enter it:
$ virtualenv $name
$ cd $name
Activate it:
$ source bin/activate
Now cd to your project directory and run the installer script:
$ cd $my_project_dir
$ python ./setup.py --prefix $path_to_virtualenv
Note the --prefix $path_to_virtualenv, which is used to tell the script to install in the virtualenv instead of system-wide. Call this after activating the virtualenv. Note that all the depencies are automatically downloaded and installed in the virtualenv.
Then you are done. When you want to leave the virtualenv, issue:
$ deactivate
On subsequent calls, you will only need to activate the virtualenv (step 2), maybe using a runawesomeproject.sh if you really want.
As noted on the virtualenv website, you should use virtualenv >= 1.9, as the previous versions did not download dependencies via HTTPS. If you consider plain HTTP to be sufficient, then any version should do.
You might also try relocatable virtualenvs: setup it and copy the folder to your host. Anyway, note that this feature is still experimental.