I am trying to set up Bazel on an existing project that consists of three applications in Python and Groovy, and a shared protobuf IDL.
For the Python applications, I currently have a custom command in setup.py that generates a Python module from the protobuf IDL. When generating the python module, I place it inside the application packages, so it can be imported like any other module in the application.
When trying to put the whole project under Bazel I'm struggeling to find out how to deal with the generated python module. protoc will only generate a single file. In order to put the file in a package, I need to create a directory structure and move the file into place. Some googling has lead me to a solution that combines some pkg_tar rules to create a tarball with the correct layout, but I can't figure out how to make the jump to making this into a python library.
The files are laid out like this:
.
├── BUILD
├── protobuf
│ └── messages.proto
└── python
└── ibidem
├── __init__.py
└── codetanks
├── __init__.py
└── domain
└── __init__.py
I want the generated module to be placed in ibidem/codetanks/domain, so that it can be imported with from ibidem.codetanks.domain import messages_pb2.
My current BUILD file:
load("#build_stack_rules_proto//python:python_proto_library.bzl", "python_proto_library")
load("#rules_pkg//:pkg.bzl", "pkg_tar", "pkg_deb")
proto_library(
name = "messages_proto",
srcs = ["protobuf/messages.proto"],
)
python_proto_library(
name = "messages_python_proto",
deps = [":messages_proto"],
)
pkg_tar(
name = "python_messages_tarball",
strip_prefix = "protobuf/",
package_dir = "ibidem/codetanks/domain",
srcs = [":messages_python_proto"],
)
filegroup(
name = "python_domain_files",
srcs = glob([
"python/**/*.py",
]),
)
pkg_tar(
name = "python_domain_tarball",
strip_prefix = "python/",
srcs = [":python_domain_files"],
)
# This fails because the tarballs doesn't have the `py` or `PyInfo` provider .
# If I use a `pkg_tar` rule here, the tarball has exactly the contents I'd want to have as a python library.
py_library(
name = "python",
deps = [
":python_domain_tarball",
":python_messages_tarball",
],
)
I've found some places that say that this can be solved by putting the messages.proto file inside the directory structure in the same place as I want the generated file to wind up. That sounds like a bad workaround, considering that placement won't make any sense for any other language than Python. I also generate a java package, and in the future the plan is to add other languages too.
Is this simply a limitation of Bazel, or can it be solved in some fancy way that I haven't been able to google my way to?
With the link provided by Sjoerd Visscher, I found a solution that seems to do the trick. First step was to separate into language specific packages. Then use copy_file to move the generated file into a subdirectory. Once that was done, combining the helper files with the generated file in a py_library was pretty straight forward.
The file layout is now:
.
├── protobuf
│ ├── BUILD
│ └── messages.proto
└── python
├── BUILD
└── ibidem
├── __init__.py
└── codetanks
├── __init__.py
└── domain
└── __init__.py
Contents of protobuf/BUILD:
package(default_visibility=["//domain:__subpackages__"])
load("#build_stack_rules_proto//python:python_proto_library.bzl", "python_proto_library")
proto_library(
name = "messages_proto",
srcs = ["messages.proto"],
)
python_proto_library(
name = "python_messages_proto",
deps = [":messages_proto"],
)
Contents of python/BUILD:
load("#build_stack_rules_proto//python:python_proto_library.bzl", "python_proto_library")
load("#rules_python//python:defs.bzl", "py_library")
load("#bazel_skylib//rules:copy_file.bzl", "copy_file")
copy_file(
name="python_messages_file",
src="//domain/protobuf:python_messages_proto",
out="ibidem/codetanks/domain/messages_pb2.py",
)
filegroup(
name = "python_helper_files",
srcs = glob([
"ibidem/**/__init__.py",
]),
)
py_library(
name = "messages",
srcs = [
":python_helper_files",
":python_messages_file",
],
visibility = ["//visibility:public"]
)
Related
I generate the documentation of my Python code from the docstrings via Doxygen (1.9.1) in addition with doxypypy (git version from today).
My project is called Project and the packages name in it (which should be imported) is mypackage. When I look into the tree sidebar of the generated html it looks like this:
└── Project
└── Packages
└── Packages
└──mypackages
The two packages are linking to the same target: ../html/namespaces.html.
The files and folders are structured like this
Project
├── LICENSE
├── README.md
├── docs
└── ...
└── src
├── mypackage
│ ├── a.py
│ ├── __init__.py
│ └── _mypackage.py
├── setup.cfg
└── setup.py
The Doxyfile is located in Project/docs, doxygen is run in there and use ../src/mypackage as INPUT directory.
More details
__init__.py
__version__ = '0.0.1a'
from ._mypackage import *
_mypackage.py
# -*- coding: utf-8 -*-
"""Example __init__.py short.
Now some longer with multiple lines. Here
comes the scond line.
"""
def foo(bar):
"""
This is foo() in mypackage.
Args:
bar (str): A paramenter.
Returns:
(int): Fixed seven.
"""
print(bar)
return 7
a.py
# -*- coding: utf-8 -*-
"""This is mypackage.a
"""
import mypackage
def bar(bar):
"""
This is the function named bar.
The function calls `mypackage.foo()` and returns an 'a'.
Paramters:
bar (str): Just a parameter.
Returns:
str: Just an 'a'.
"""
mypackage.foo(bar)
return('a')
Some (maybe) related Doxyfile settings
BRIEF_MEMBER_DESC = YES
REPEAT_BRIEF = YES
ALWAYS_DETAILED_SEC = NO
FULL_PATH_NAMES = YES
JAVADOC_AUTOBRIEF = NO
PYTHON_DOCSTRING = YES
OPTIMIZE_OUTPUT_FOR_C = NO
OPTIMIZE_OUTPUT_JAVA = YES
OPTIMIZE_FOR_FORTRAN = NO
OPTIMIZE_OUTPUT_VHDL = NO
OPTIMIZE_OUTPUT_SLICE = NO
MARKDOWN_SUPPORT = YES
EXTRACT_ALL = YES
EXTRACT_PRIVATE = YES
EXTRACT_PACKAGE = YES
EXTRACT_STATIC = YES
EXTRACT_LOCAL_CLASSES = YES
EXTRACT_LOCAL_METHODS = YES
EXTRACT_ANON_NSPACES = NO
RESOLVE_UNNAMED_PARAMS = YES
HIDE_UNDOC_MEMBERS = NO
HIDE_UNDOC_CLASSES = NO
HIDE_FRIEND_COMPOUNDS = NO
HIDE_IN_BODY_DOCS = NO
INPUT = ../src/mypackage
FILE_PATTERNS =
RECURSIVE = YES
FILTER_PATTERNS = *.py=./py_filter
GENERATE_HTML = YES
GENERATE_TREEVIEW = YES
I opened an Issue about that.
This question is exactly like this stackoverflow question. And I am hoping I get some different answer with this one. I have been trying to achieve this for over a year. but can't seem to achieve it.
Stripped down version of my ansible directory looks like this:
[root#python-test ansible]# tree
.
├── files
├── inventory
│ ├── prod
│ │ ├── group_vars
│ │ │ └── all
│ │ └── hosts -> ../../scripts/inventory.py
│ └── staging
│ ├── group_vars
│ │ └── all
│ └── hosts -> ../../scripts/inventory.py
├── roles
│ ├── ant
│ ├── build
│ ├── jdk
│ └── python
├── scripts
│ └── inventory.py
├── templates
└── vars
└── all.yaml
I would like to use some of the variables declared in group_vars/all file. It has some endpoint details that I can use during the script execution. A striped down version of this group_vars/all looks like this:
inv_cloudprovider: aws
inv_environment: prod
inv_environment_type: production
inv_vpc_cidr: 10.0.0.0/16
inv_build_url: 'https://build.{{inv_environment}}.local'
inv_cloud_vpc_name: '{{inv_environment}}-vpc'
inv_vpc_id: '{{inv_environment_type}}-{{inv_cloud_vpc_name}}'
inv_glassfish_version: 4
At this point I am loading this file using yaml.safe_loads() and then using them in script execution. but problem with that is there are variables which are recursive jinja templates. and I am having hard time getting to make those variables work. So I was wondering if I can use ansible to do this for me.
I am using ansible==2.2.3.0 and python 2.7.
The closest I have been to achieving this in python by doing this:
from ansible.parsing.dataloader import DataLoader
from ansible.vars import VariableManager
from ansible.inventory import Inventory
inventory_path = '/root/ansible/inventory/prod/hosts'
inventory = Inventory(DataLoader(), VariableManager(), inventory_path)
group = inventory.get_group('all')
group_vars = group.get_vars()
Which is probably not the correct way because my script tries to execute itself and that goes on recursively.
Is it possible to parse group vars variable using ansible? If no, how do I best get final value of variables from that file?
The stripped down version of code that can do what this question asked is:
#!/usr/bin/env python
import json
import os
import sys
import yaml
from jinja2 import Environment
class VarLoader(yaml.SafeLoader):
#staticmethod
def construct_python_string(text):
return str(text.value)
def yaml_remove_parse(self, text):
self.construct_python_string(text)
env_name = os.environ.get('ENV_NAME')
script_path = os.path.dirname(os.path.abspath(__file__))
ansible_dir_path = os.path.abspath(script_path + '/../')
global_vars_path = ansible_dir_path + '/vars/all.yaml'
inventory_vars_path = ansible_dir_path + '/inventory/' + env_name + '/group_vars/all'
with open(global_vars_path) as f1, open(inventory_vars_path) as f2:
global_vars_data = f1.read()
inventory_vars_data = f2.read()
all_vars_data = global_vars_data + '\n' + inventory_vars_data
VarLoader.add_constructor('tag:yaml.org,2002:float', VarLoader.yaml_remove_parse)
data = yaml.load(all_vars_data, Loader=VarLoader)
jinja_env = Environment()
template_file = jinja_env.from_string(all_vars_data)
for key in data:
val = data[key]
if isinstance(val, str) and '{{' in val:
template = jinja_env.from_string(val)
data[key] = template.render(**data)
yaml_data = yaml.load(template_file.render(data), Loader=VarLoader)
print(json.dumps(yaml_data))
This works with both python3 and python2. Output:
[root#python-test ansible]# python scripts/inventory.py | jq .
{
"inv_environment": "prod",
"inv_environment_type": "production",
"inv_cloudprovider": "aws",
"gv_redis_port": 6379,
"inv_glassfish_version": 4,
"inv_vpc_id": "production-prod-vpc",
"gv_redis_version": "4.0.11",
"inv_build_url": "https://build.prod.local",
"gv_glassfish_admin_user": "admin",
"inv_vpc_cidr": "10.0.0.0/16",
"gv_glassfish_asadmin_path": "/usr/local/glassfish/bin/asadmin",
"gv_java_home": "/usr/local/java/default",
"inv_cloud_vpc_name": "prod-vpc",
"gv_tomcat_home": "/usr/local/tomcat",
"gv_java_path": "/usr/local/java/default/bin/java"
}
Suppose we have a project like this:
project-path
├── root
│ ├── BUILD
│ ├── gen
│ │ ├── a2.txt
│ │ └── a.txt
│ └── use.py
└── WORKSPACE
And in use.py:
f = open("gen/a.txt", "r")
f2 = open("gen/a2.txt", "r")
print(f.read())
print(f2.read())
And BUILD:
py_binary(
name = "use",
srcs = ["use.py"],
data = ["gen/a.txt", "gen/a2.txt"],
)
when I bazel run root:use, it errors:
FileNotFoundError: [Errno 2] No such file or directory: 'gen/a.txt'
It expects paths relative to the WORKSPACE directory, not the current package (root/gen/a.txt here). But I want to access files relative to each package.
Add Skylib to your project, i.e. extend you WORKSPACE file:
WORKSPACE
load("#bazel_tools//tools/build_defs/repo:http.bzl", "http_archive")
http_archive(
name = "bazel_skylib",
sha256 = "74d544d96f4a5bb630d465ca8bbcfe231e3594e5aae57e1edbf17a6eb3ca2506",
urls = [
"https://mirror.bazel.build/github.com/bazelbuild/bazel-skylib/releases/download/1.3.0/bazel-skylib-1.3.0.tar.gz",
"https://github.com/bazelbuild/bazel-skylib/releases/download/1.3.0/bazel-skylib-1.3.0.tar.gz",
],
)
load("#bazel_skylib//:workspace.bzl", "bazel_skylib_workspace")
bazel_skylib_workspace()
Replace the path in the bash script using text replacement:
load("#bazel_skylib//rules:expand_template.bzl", "expand_template")
expand_template(
name = "modifiy_use_for_bazel",
out = "prepared_for_bazel_use.py",
substitutions = {
"gen/a.txt": "root/gen/a.txt",
"gen/a2.txt": "root/gen/a.txt",
},
template = "use.py",
)
py_binary(
name = "use",
main = "prepared_for_bazel_use.py",
srcs = ["prepared_for_bazel_use.py"],
data = [
"gen/a.txt",
"gen/a2.txt",
],
)
You can run now the script via bazel run root:use
Note: Tested with Bazel 6.0.0
To make this all a bit more convenient to use you could implement your own rule for it since this seems to be a common problem (e.g. when supporting two build systems at the same time where Bazel is not the primary build system and the other build system can cope with relative path names.)
A similar problem is described here: Change test execution directory in Bazel?
I cannot eliminate the doctest flags (ie. <BLANKLINE>, # doctest: +ELLIPSIS) for the html output. I am able to generate the documentation as I would like, so no errors there but it includes theses flags which I would like removed. Sphinx documentation here claims this is possible so I must be doing something wrong. My documentation examples are in numpy style and I have tried using both the napoleon and numpydoc extensions.
Here are the steps I have taken.
run sphinx-quickstart (enabling autodoc and doctest extensions)
run sphinx-apidoc to generate .rst files
run make doctest (all tests are passing)
run make html
I have tried the setting trim_doctest_flags and doctest_test_doctest_blocks variables in conf.py with no success.
Is there something I am missing to trigger sphinx to remove these for the html docs? I am hoping this is enough information to get pointed in the right direction since the docs look good except for this one issue. However, I can provide more details or an example if necessary.
Update: MCV Example (Using Sphinx 1.8.2)
directory and file structure
.
├── trial
│ ├── __init__.py
│ └── trial.py
└── trialDocs
├── build
├── Makefile
└── source
├── _static
├── _templates
├── conf.py
├── index.rst
├── modules.rst
└── trial.rst
conf.py
# -*- coding: utf-8 -*-
#
# Configuration file for the Sphinx documentation builder.
import os
import sys
sys.path.insert(0, os.path.abspath('../../trial'))
project = 'trial'
copyright = '2019, trial'
author = 'trial'
version = ''
release = 'trial'
extensions = [
'sphinx.ext.autodoc',
'sphinx.ext.doctest',
'sphinx.ext.napoleon',
]
templates_path = ['_templates']
source_suffix = '.rst'
master_doc = 'index'
language = None
exclude_patterns = []
pygments_style = None
html_theme = 'alabaster'
htmlhelp_basename = 'trialdoc'
latex_elements = {}
latex_documents = [(master_doc, 'trial.tex', 'trial Documentation', 'trial', 'manual'),]
man_pages = [(master_doc, 'trial', 'trial Documentation', [author], 1)]
texinfo_documents = [(master_doc, 'trial', 'trial Documentation', author, 'trial', 'One line description of project.', 'Miscellaneous'),]
epub_title = project
epub_exclude_files = ['search.html']
doctest_global_setup = """
from trial import *
"""
trim_doctest_flags=True
trial.rst - this was generated using sphinx-apidoc
trial package
=============
Module contents
---------------
.. automodule:: trial
:members:
:undoc-members:
:show-inheritance:
trial.py
def withBlankline():
"""
Use blanklines in example.
Determine if sphinx will eliminate <BLANKLINE> for html.
Examples
--------
>>> withBlankline()
<BLANKLINE>
blanklines above and below
<BLANKLINE>
"""
print()
print('blanklines above and below')
print()
class Example():
def __init__(self):
pass
def withEllipsis(self):
"""
Use ellipsis in example.
Determine if sphinx will eliminate # doctest: +ELLIPSIS for html.
Examples
--------
>>> e = Example()
>>> e.withEllipsis() # doctest: +ELLIPSIS
abc...xyz
"""
print('abcdefghijklmnopqrstuvwxyz')
using make html or sphinx-build -b html source build
trial.html output:
Based on the comment by #mzjn, it appears that this was a bug, fixed in Sphinx 2.2.0:
Issue: doctest comments not getting trimmed since Sphinx 1.8.0 - Issue #6545
I have a flask project with the following structure:
├─ app.py
├─ project
| ├─ __init__.py
| └─ static
| └─ typescript
| └─ app.ts
└─ typings
├─ globals
| └─ ... # multiple imported ts libraries
└─ index.d.ts
I'm using a webpacker integration called Flask Assets. I've set up the compilation like so (in __init__.py)
ts = get_filter('typescript')
ts.load_paths = [
#os.path.join(config.APP_ROOT, '..', 'typings'), # doesn't do anything :/
os.path.join(app.static_folder, 'typescript')
]
assets.register('javascript', Bundle(
'typescript/app.ts',
filters = (ts, 'jsmin'),
output = 'js/app-%(version)s.js'
))
My app.ts is, more or less,
class SomeClass {
... various class methods, using things like jQuery and CryptoJS
}
no imports - I'm not really sure whether or not I need them.
The specific error I'm getting is
Cannot find name 'JQuery'.
../../../../../var/folders/5t/4x0gmsdx0dbbgv_fr3cv3x6m0000gn/T/tmphFTSQo.ts(7,17): error TS2503: Cannot find namespace 'CryptoJS'.
../../../../../var/folders/5t/4x0gmsdx0dbbgv_fr3cv3x6m0000gn/T/tmphFTSQo.ts(10,27): error TS2304: Cannot find name '$'.
... a bunch more about other symbols
I kind of solved it...
glob_string = os.path.join(config.APP_ROOT, '..', 'typings', '*', '*', '*.d.ts')
assets.register('javascript', Bundle(
glob.glob(glob_string),
'typescript/app.ts',
filters = ('typescript', 'jsmin'),
output = 'js/app-%(version)s.js'
))
basically I just "manually" add all the definition files to the bundle (using glob). It's not sufficient to just add the index.d.ts in the root of the typings dir as the typescript filter copies the .ts to a temp file (in /tmp) before compiling and the paths in index.d.ts are relative.
it should also be noted that ts.load_paths does nothing...