goto_definition must ignore import lines and return only real definitions.

When i decided to try this package out today, I ran into a little bit of trouble. Wasn't until I removed the filter on the server in the elisp file I understood that the problem was I didn't have pip or python-json-rpc installed.

Perhaps python-json-rpc could be pointed out as a dependency on the front page (and also the pip-behavior)?

That way, I can just use "q" instead of "C-x 1" to get rid of the doc buffer

I am using company-anaconda and anaconda from MELPA

Thanks,

Hi,

Anaconda mode looks very interesting. Could you give some motivation as to why a user should prefer it to Elpy (https://github.com/jorgenschaefer/elpy), which I am currently using. What are your unique strengths? To my untrained eye it looks like you cover very similar ground.

Thanks

P.S. I discovered you via a link from the jedi docs: http://jedi.jedidjah.ch/en/latest/

Some times silently anaconda_mode fail to start and doesn't say user anything. Most times this happens because specified port already in use. Any way its necessary raise a error.

In anaconda-mode-python, python executable is hardcoded as "python" . This is not convinient, especially when you have python3 and python2 installed at the same time. I think it will be better if you change it to use python-shell-interpreter.
Anyway, thank you for this great mode. I like it.

when open two emacs instances editing different python files, the emacs (actually the computer) will hang. The CPU is 100% used.

see redguardtoo/emacs.d#133

As requested in 984050a#commitcomment-9329398, here is the traceback when I edit (triggering completion) a Python file.

Debugger entered--Lisp error: (error "Completion needs an inferior Python process running")
  signal(error ("Completion needs an inferior Python process running"))
  error("Completion needs an inferior Python process running")
  python-completion-complete-at-point()
  completion--capf-wrapper(python-completion-complete-at-point optimist)
  run-hook-wrapped(completion--capf-wrapper python-completion-complete-at-point optimist)
  company--capf-data-real()
  company--capf-data()
  company-capf(prefix)
  #[0 "\n\203   \nC\202\n�\211\205\245\211@\300�9\204�\306�!\203<�9\2036�\307N\310=\2046�\307N?\205@\311�!\205@�\312!\202@\313�\312\"\240\210\300\242\203\236\314\300\242!\203\230\315\300\242!�\211�\301\316\f!\240\210\301\242:\204l
\203\230\317\320!\210\202\230
\203t\300\242��\2119\203\203\321\322�!P��\202\211\323\301\242@!\210\324\301\242!\210\325\326\327 \"\210\330\331!\210\332\333\301\242\"\210�A\266\202\202\n\207" [(nil) (nil) company-backend company-backends company-prefix company--manual-action functionp company-init t company-init-backend prefix company--multi-backend-adapter company--good-prefix-p company--prefix-str company-calculate-candidates message "No completion found" " " symbol-name company--update-group-lighter company-update-candidates run-hook-with-args company-completion-started-hook company-explicit-action-p company-call-frontends show throw --cl-block-nil-- company--manual-prefix company-lighter] 6 "\n\n(fn)"]()
  funcall(#[0 "\n\203   \nC\202\n�\211\205\245\211@\300�9\204�\306�!\203<�9\2036�\307N\310=\2046�\307N?\205@\311�!\205@�\312!\202@\313�\312\"\240\210\300\242\203\236\314\300\242!\203\230\315\300\242!�\211�\301\316\f!\240\210\301\242:\204l
\203\230\317\320!\210\202\230
\203t\300\242��\2119\203\203\321\322�!P��\202\211\323\301\242@!\210\324\301\242!\210\325\326\327 \"\210\330\331!\210\332\333\301\242\"\210�A\266\202\202\n\207" [(nil) (nil) company-backend company-backends company-prefix company--manual-action functionp company-init t company-init-backend prefix company--multi-backend-adapter company--good-prefix-p company--prefix-str company-calculate-candidates message "No completion found" " " symbol-name company--update-group-lighter company-update-candidates run-hook-with-args company-completion-started-hook company-explicit-action-p company-call-frontends show throw --cl-block-nil-- company--manual-prefix company-lighter] 6 "\n\n(fn)"])
  company--begin-new()
  company--perform()
  #[0 "\301 \210�\207" [company-candidates company--perform] 1 "\n\n(fn)"]()
  funcall(#[0 "\301 \210�\207" [company-candidates company--perform] 1 "\n\n(fn)"])
  company-auto-begin()
  company-idle-begin(#<buffer api.py> #<window 8 on api.py> 756 814)
  apply(company-idle-begin (#<buffer api.py> #<window 8 on api.py> 756 814))
  byte-code("r\301�\302H�\303H\"\210)\301\207" [timer apply 5 6] 4)
  timer-event-handler([t 21691 56161 440532 nil company-idle-begin (#<buffer api.py> #<window 8 on api.py> 756 814) nil 0])

Additionally, as I enter the Python file I have this message:

Compiler-macro error for python-syntax-context: (void-function python-syntax--context-compiler-macro) [2 times]

Thanks : )

I'm on Emacs 24.3.50.1 and have installed the latest anaconda-mode from Melpa.

Hey I have my http_proxy envs set which seems to be causing issues with completion

Using a proxy for http...                                                                        
Making url-show-status local to  *http 127.0.0.1:3128* while let-bound!                          
anaconda-mode-view-doc: Can't find documentation                                                 
Missing module docstring (C0111)                                                                 
Beginning of buffer [2 times]                                                                    
Mark set                                                                                         
Company: An error occurred in auto-begin                                                         
Company: Back-end company-anaconda error "End of file during parsing" with args (candidates fire)
Company: An error occurred in auto-begin                                                         
Company: Back-end company-anaconda error "End of file during parsing" with args (candidates fire)
Making completion list...                                                                        

There are following contib extensions planing by default in 0.2.0 version:

• anaconda-mode-company-mode (company mode will be extension due to project renaming)
• anaconda-mode-helm-mode (add automatically if helm available)
• anaconda-mode-imenu-mode

One little question. I'm using company-anaconda mode. And I'm getting some completions really good.
For example for:

import numpy as np
np.

After the . I'm getting good completions. But with some other libraries I don't. I'm sure that it's because of the libraries being loaded and the default paths. How can I add more paths for this usage?
Or is it a company-anaconda question?

Trying to install this package from MELPA, emacs version: GNU Emacs 24.3.1 (i386-mingw-nt6.1.7600)

Compile log:

Leaving directory `c:/Users/dmishin/AppData/Roaming/.emacs.d/elpa/anaconda-mode-20140327.142'

Compiling file c:/Users/dmishin/AppData/Roaming/.emacs.d/elpa/anaconda-mode-20140327.142/anaconda-eldoc.el at Thu Mar 27 13:32:33 2014
Entering directory `c:/Users/dmishin/AppData/Roaming/.emacs.d/elpa/anaconda-mode-20140327.142/'

Compiling file c:/Users/dmishin/AppData/Roaming/.emacs.d/elpa/anaconda-mode-20140327.142/anaconda-mode-pkg.el at Thu Mar 27 13:32:33 2014

Compiling file c:/Users/dmishin/AppData/Roaming/.emacs.d/elpa/anaconda-mode-20140327.142/anaconda-mode.el at Thu Mar 27 13:32:33 2014

In anaconda-mode-do-request:
anaconda-mode.el:109:26:Warning: reference to free variable
    `url-http-response-status'

Compiling file c:/Users/dmishin/AppData/Roaming/.emacs.d/elpa/anaconda-mode-20140327.142/company-anaconda.el at Thu Mar 27 13:32:33 2014
company-anaconda.el:26:1:Error: Cannot open load file: company

Hello,
I have a jedi-mode, autocomplete, projectile setup right now.
And I want to change it to anaconda-mode, company-mode, projectile.
When I checked the code I saw that anaconda-mode has hardcoded integration to cask. But I want to use projectile instead. Is it possible to use projectile with anaconda-mode? I want to start anaconda server on projectile project root. So goto-source commands would work right.
Here is how I change the root in my current setup.
https://github.com/huseyinyilmaz/emacs-config/blob/master/lisp/user-init-jedi.el#L52-L56

When I do M-. it open the new buffer or position in the same window. Would be great if with C-u M-. it open the new buffer in the other window, like in find-file-other-window

So remote autocomplete is really cool and I plan to add it to anaconda-mode. So this is check list to implement this feature.

• this feature considered to be enable if anaconda-mode-host set to 'vagrant symbol or doesn't set to loop back interface
• start anaconda-mode server with ssh tunnel
• check that anaconda mode alive by checking ssh connection
• in case of tramp buffers translate filename to local file name of remote machine

http://val.markovic.io/articles/youcompleteme-as-a-server

https://github.com/Valloric/ycmd

I assume that the architecture is pretty similar to communicating with jedi, even simpler. Maybe transfer anaconda-mode's code to that is the fastest way to get it implemented.

Just in case you're interested :)

Hi, thanks for this great and simple plugin.

I am developing some private modules in emacs with anaconda-mode. I want anaconda to aware of the modules I am developing.
I need to edit sys.path to inform anaconda about this module, is there a configuration to allow me do that?

Best,

Few first requests will always fail due to server initialization lag. Server must be started synchronously with emacs. Protect anaconda-mode-call with and function. This'll allow don't perform rpc calls on broken nodes.

company-jedi-add-python-path command allow user include special directories to python path throw read-dir-name

Hello - goto definition works great. I just can't figure this out, but how would I lookup a definition by searching for it by typing the name ? maybe integrated with helm.

In Atom or Sublime its easy to do cmd-shift-R, lookup the class name and go straight to the definition.

I've made an etags file for emacs, but its matching imports and usages. When I search in helm it doesn't know where the class def itself is.

Obviously jedi knows where the class def is (because goto definition works correctly)

So is this a command that anaconda-mode should implement ? Or should I look more at helm ?

Thanks for your help and for the package.

Find usages result must return code snippet not line number duplication. The good point to fix this is jedi it self.

https://github.com/pavlov99/json-rpc
https://github.com/skeeto/elisp-json-rpc

Use python-shell-virtualenv-path variable for python executable search. Restart anaconda_mode if current environment different with current node instance.

Since now anaconda use pip to install jedi, using assert to check version of jedi and jsonrpc may fail on some systems. In my case, I use mac os x, and pip1.5.6, my jedi version is "0.8.1-final10". This causes error in anaconda_mode.py.
I fixed it on my computer by modify line 37 of anaconda_mode.py:

assert jedi.__version__ == '0.8.1-final0', 'Jedi version does not match 0.8.1'

But obviously this is not a good solution.
Please fix this. Thanks.

I've downloaded and installed company-jedi with cask and MELPA, running Emacs 24.3 on Arch Linux. Because I already have jedi installed for python3, I've used (setq company-jedi-bin "python3") before loading company-jedi with (require 'company-jedi). I switch to a sample python file looking like this (| is point):

def main():
    pas|

if __name__ == '__main__':
    main()

I start the company-jedi-process with (company-jedi-start). No complaints there. htop tells me it's running. I execute (company-jedi-candidates) and it returns nil. I do a counter-check by opening a python3 instance and making a sample request with the requests module, using the output pasted from (company-jedi-request-json "candidates"):

>>> import requests
>>> requests.post('http://localhost:24970', data="{\"command\":\"candidates\", \"attributes\":{\"source\":\"def main():\\n    pas\\n\\nif __name__ == '__main__':\\n    main()\\n\", \"line\":2, \"column\":7, \"point\":19, \"path\":\"\\/home\\/wasa\\/tmp\\/complete.py\", \"company_prefix\":\"\", \"company_arg\":\"\"}}").text
'["pass"]'

That proves the server is running and is able to return the correct response. I suspect something goes wrong when doing the POST request, simply because (with-current-buffer (company-jedi-retrive (company-jedi-request-json "candidates")) (buffer-string)) returns an empty string. Could you investigate this?

Please see #49 for an explanation; restoring the line sys.path.insert(0, '') resolves the issue.

Select definitions and references similar to helm-swoop.

Starting the HTTP server on '0.0.0.0' is a security concern for those running on a machine without a properly configured firewall. This fix changes default ip interface to 'localhost', and allow override from command line by '--ip' option.

diff --git a/company-jedi.el b/company-jedi.el
index 6620529..2014eb8 100644
--- a/company-jedi.el
+++ b/company-jedi.el
@@ -64,6 +64,7 @@
   (mapconcat 'identity
              (list company-jedi-python-bin "-m" "start_jedi"
                    "--port" (number-to-string company-jedi-port)
+                   "--ip" company-jedi-host
                    (when company-jedi-debug "--debug"))
              " ")
   "Command to run start_jedi server.")
diff --git a/start_jedi/__init__.py b/start_jedi/__init__.py
index c028e80..02e1b87 100644
--- a/start_jedi/__init__.py
+++ b/start_jedi/__init__.py
@@ -2,10 +2,10 @@ import start_jedi.httpd
 import start_jedi.cli


-def run(port):
+def run(port, ip='127.0.0.1'):
     """Run jedi server."""

-    address = ('', port)
+    address = (ip, port)

     handler = start_jedi.httpd.JediHandler

diff --git a/start_jedi/__main__.py b/start_jedi/__main__.py
index 1ae2b6c..988db8e 100644
--- a/start_jedi/__main__.py
+++ b/start_jedi/__main__.py
@@ -22,4 +22,4 @@ if args.debug:
     logger.addHandler(handler)


-start_jedi.run(args.port)
+start_jedi.run(args.port, args.ip)
diff --git a/start_jedi/cli.py b/start_jedi/cli.py
index 1d8ea3b..97347a6 100644
--- a/start_jedi/cli.py
+++ b/start_jedi/cli.py
@@ -10,6 +10,8 @@ def args():
     parser.add_argument('-d', '--debug', action='store_true',
                         help='enable debug logging')

+    parser.add_argument('-i', '--ip', type=str, default='localhost',
+                        help='server IP')
     parser.add_argument('-p', '--port', type=int, default=8000,
                         help='server port number')

I'm using company-anaconda, though I don't think it's related to that package. Here's the JSON-RPC error that Company gets back:

JSON-RPC error condition: :data, (:type "AttributeError\" :args ("'FakeStatement' object has no attribute 'path'") :message "'FakeStatement' object has no attribute 'path'"), :code, -32000, :message, "Server error"" with args (candidates )"))

Jedi provides goto_definitions, goto_assignments and usages that each returns a list of definitions. I think we need a mode for navigating such a list. Below is the usages of script_method. Definitions are grouped by module, inspired by occur.

You navigate up and down with n and p, and a preview is shown in other-window. If you press enter or click a row, you jump to that location. When you jump, the current position is pushed to a stack so that you are able to navigate back.

goto_definitions and goto_assignments should immediately jump if there is just 1 candidate. M-. should be bound to goto_definitions, and we can have another key for goto_assignments. I don't think goto_definitions and goto_assignments should be mixed.

What do you think @proofit404, is this a good idea? I have implemented some parts of this.

Code smell.

If anaconda-mode-start is called in non python mode we must ignore anaconda-mode-start-stop-hook run.

Rebind find-tag-other-window and find-tag-other-frame to equivalent commands go to definition and find reference functionality. Read definition name from package module to specified method if no symbol at point available or C-u given. This behaviour must be provided in anaconda-mode-view-doc function too.

For provide interpreter support we must obtain local() namespace from it. This require extra interaction with inferior process similar to slime. This is out of anaconda-mode aims.

Hi,

I just want to know if there a way to avoid the "active process exist; kill then and exit anyway?" when I close Emacs and there is the running process python anaconda_mode.py. If it's possible to configure it, do you know how I can do? If not, do you think it's possible to slightly modify the code?

I quickly tried jedi and there is no such message. I switch to anaconda because I replace auto-complete to company and jedi depends on auto-complete.

Thanks,
Damien

Hi--

I've been playing around for an hour with anaconda-mode, but I can't seem to get it to work.

I've installed the json-rpc and jedi from pip. I didn't get a specific version like in the req file, I just used pip install jedi etc. I've installed anaconda-mode and pyenv-mode from melpa and added the following to my .emacs:

;;trying to get anaconda-mode to work
(pyenv-mode)
(add-hook 'python-mode-hook 'anaconda-mode)

I now have a json-rpc process running in the background and I can choose stuff like anaconda-mode when I press M-x. Still the various anaconda-mode options do not seem to do anything and I do not get any autocompletions.

Can you tell what I am doing wrong? Have I missed any crucial steps? Perhaps I also need the jedi.el package?

(And it might be nice for other newbs if you added some minimal steps needed to install conda mode and what you should expect to see if you have installed it correctly. Seems like a really useful package this, would be a shame if people are prevented from using due to it being slightly hard to install.)

Define company-jedi-work-on, company-jedi-deactivate and company-jedi-virtualenv commands in emacs lisp side. This commands will use /virtualenv/ resource of start_jedi api. After activation of some virtualenv its name appears in the mode line after vc branch name. Something like Env:django-env. Ask user to switch virtualenv automatically if current file changes on the disk. This happens mostly on switching vcs branch and many projects use different virtualenvs in different branches.

I get this when I try to use anaconda-go-to-definition :( Can you guide me on what information should I give to solve this? Running Emacs 24.4 Ubuntu 14.04

Question: does the name anaconda has anything to do with the Python distribution Anaconda and its tool conda?

Add following sections to README:

• customization section with code snippets for users configs
• explain keybindings
• real world example for django project with branch switching
• how its work

Hello,
I was having some problems with the installation and in the end I tried a simple command python anaconda_mode.py for which I got the following result:

$ python  anaconda_mode.py
Traceback (most recent call last):
  File "/home/atlas/.emacs.d/elisp/anaconda-mode/anaconda_mode.py", line 2, in <module>
    from jsonrpc import dispatcher, JSONRPCResponseManager
ImportError: cannot import name dispatcher

I just installed jsonrpc with sudo pip install jsonrpc. Have you any idea what the problem could be?

Use anaconda-mode to populate imenu.

I've had this issue recently after updating anaconda-mode from melpa. My emacs hung when company-mode tries to auto complete with anaconda, and I see message as following in minibar.

Blocking call to accept-process-output with quit inhibited.

I guess this is a regression as I saw previous discussion on this issue before. Below is backtrace log on toggle-debug-on-quit.

Debugger entered--Lisp error: (quit)
  anaconda-mode-bootstrap()
  anaconda-mode-start()
  anaconda-mode-call("complete")
  anaconda-mode-complete()
  company-anaconda-candidates()
  company-anaconda(candidates #("fo" 0 2 (fontified nil)))
  apply(company-anaconda (candidates #("fo" 0 2 (fontified nil))))
  #[0 "\302 !\203��\303 \300\"\207\303\304  \300#\207" [(candidates #("fo" 0 2 (fontified nil))) company-backend functionp apply company--multi-backend-adapter] 4 "\n\n(fn)"]()
  funcall(#[0 "\302 !\203��\303 \300\"\207\303\304  \300#\207" [(candidates #("fo" 0 2 (fontified nil))) company-backend functionp apply company--multi-backend-adapter] 4 "\n\n(fn)"])
  company-call-backend-raw(candidates #("fo" 0 2 (fontified nil)))
  company--fetch-candidates(#("fo" 0 2 (fontified nil)))
  company-calculate-candidates(#("fo" 0 2 (fontified nil)))
  #[0 "\n\203   �\nC\202\n��\211\205\245�\211@\300�9\204��\306�!\203<��9\2036��\307N\310=\2046��\307N?\205@�\311�!\205@��\312!\202@�\313�\312\"\240\210\300\242\203\236�\314\300\242!\203\230�\315\300\242!�\211�\301\316\f!\240\210\301\242:\204l�
\203\230�\317\320!\210\202\230�
\203t�\300\242��\2119\203\203�\321\322�!P��\202\211�\323\301\242@!\210\324\301\242!\210\325\326\327 \"\210\330\331!\210\332\333\301\242\"\210�A\266\202\202\n�\207" [(#("fo" 0 2 (fontified nil))) (nil) company-backend company-backends company-prefix company--manual-action functionp company-init t company-init-backend prefix company--multi-backend-adapter company--good-prefix-p company--prefix-str company-calculate-candidates message "No completion found" " " symbol-name company--update-group-lighter company-update-candidates run-hook-with-args company-completion-started-hook company-explicit-action-p company-call-frontends show throw --cl-block-nil-- company--manual-prefix company-lighter] 6 "\n\n(fn)"]()
  funcall(#[0 "\n\203   �\nC\202\n��\211\205\245�\211@\300�9\204��\306�!\203<��9\2036��\307N\310=\2046��\307N?\205@�\311�!\205@��\312!\202@�\313�\312\"\240\210\300\242\203\236�\314\300\242!\203\230�\315\300\242!�\211�\301\316\f!\240\210\301\242:\204l�
\203\230�\317\320!\210\202\230�
\203t�\300\242��\2119\203\203�\321\322�!P��\202\211�\323\301\242@!\210\324\301\242!\210\325\326\327 \"\210\330\331!\210\332\333\301\242\"\210�A\266\202\202\n�\207" [(#("fo" 0 2 (fontified nil))) (nil) company-backend company-backends company-prefix company--manual-action functionp company-init t company-init-backend prefix company--multi-backend-adapter company--good-prefix-p company--prefix-str company-calculate-candidates message "No completion found" " " symbol-name company--update-group-lighter company-update-candidates run-hook-with-args company-completion-started-hook company-explicit-action-p company-call-frontends show throw --cl-block-nil-- company--manual-prefix company-lighter] 6 "\n\n(fn)"])
  company--begin-new()
  company--perform()
  #[0 "\300 \207" [company--perform] 1 "\n\n(fn)"]()
  funcall(#[0 "\300 \207" [company--perform] 1 "\n\n(fn)"])
  company-auto-begin()
  company-idle-begin(#<buffer *scratch*> #<window 3 on *scratch*> 5 3)
  apply(company-idle-begin (#<buffer *scratch*> #<window 3 on *scratch*> 5 3))
  byte-code("r\301�\302H�\303H\"\210)\301\207" [timer apply 5 6] 4)
  timer-event-handler([t 21662 39799 656946 nil company-idle-begin (#<buffer *scratch*> #<window 3 on *scratch*> 5 3) nil 0])

The OS I used is Ubuntu 14.04, and the version of Emacs is 24.3.1.
My emacs config files is copied from https://github.com/redguardtoo/emacs.d, which using anaconda-mode for python scripts complement.
I opened first emacs program to edit python file in ubuntu terminal with emacs -nw a.py, and opened second emacs program to edit python file in another ubuntu terminal with emacs -nw b.py. The first emacs worked fine, but the second emacs has some problem. The emacs was working in vim mode, and when I press 'j' to move the cursor in the second emacs, I got some mseeage in message buffer:
Blocking call to accept-process-output with quit inhibited!!
Company back-end 'company-jedi' could not be initialized:
Symbol's function definition is void: company-jedi

then, the second emacs stuck, no response for key operations, and CPU usage was up to 100%, I had to close terminal to kill the second emacs.