Batch Code Processor

For an introduction and motivation to using batch processors, see here. This page describes the Batch-code processor. The Batch-command one is covered here.

Basic Usage

The batch-code processor is a superuser-only function, invoked by

> @batchcode path.to.batchcodefile

Where path.to.batchcodefile is the path to a batch-code file. Such a file should have a name ending in “.py” (but you shouldn’t include that in the path). The path is given like a python path relative to a folder you define to hold your batch files, set by BATCH_IMPORT_PATH in your settings. Default folder is (assuming your game is called “mygame”) mygame/world/. So if you want to run the example batch file in mygame/world/batch_code.py, you could simply use

> @batchcode batch_code

This will try to run through the entire batch file in one go. For more gradual, interactive control you can use the /interactive switch. The switch /debug will put the processor in debug mode. Read below for more info.

The batch file

A batch-code file is a normal Python file. The difference is that since the batch processor loads and executes the file rather than importing it, you can reliably update the file, then call it again, over and over and see your changes without needing to @reload the server. This makes for easy testing. In the batch-code file you have also access to the following global variables:

  • caller - This is a reference to the object running the batchprocessor.
  • DEBUG - This is a boolean that lets you determine if this file is currently being run in debug-mode or not. See below how this can be useful.

Running a plain Python file through the processor will just execute the file from beginning to end. If you want to get more control over the execution you can use the processor’s interactive mode. This runs certain code blocks on their own, rerunning only that part until you are happy with it. In order to do this you need to add special markers to your file to divide it up into smaller chunks. These take the form of comments, so the file remains valid Python.

Here are the rules of syntax of the batch-code *.py file.

  • #CODE as the first on a line marks the start of a code block. It will last until the beginning of another marker or the end of the file. Code blocks contain functional python code. Each #CODE block will be run in complete isolation from other parts of the file, so make sure it’s self-contained.
  • #HEADER as the first on a line marks the start of a header block. It lasts until the next marker or the end of the file. This is intended to hold imports and variables you will need for all other blocks .All python code defined in a header block will always be inserted at the top of every #CODE blocks in the file. You may have more than one #HEADER block, but that is equivalent to having one big one. Note that you can’t exchange data between code blocks, so editing a header-variable in one code block won’t affect that variable in any other code block!
  • #INSERT path.to.file will insert another batchcode (Python) file at that position.
  • A # that is not starting a #HEADER, #CODE or #INSERT instruction is considered a comment.
  • Inside a block, normal Python syntax rules apply. For the sake of indentation, each block acts as a separate python module.

Below is a version of the example file found in evennia/contrib/tutorial_examples/.

#
# This is an example batch-code build file for Evennia.
#

#HEADER

# This will be included in all other #CODE blocks

from evennia import create_object, search_object
from evennia.contrib.tutorial_examples import red_button
from typeclasses.objects import Object

limbo = search_object('Limbo')[0]


#CODE

red_button = create_object(red_button.RedButton, key="Red button",
                           location=limbo, aliases=["button"])

# caller points to the one running the script
caller.msg("A red button was created.")

# importing more code from another batch-code file
#INSERT batch_code_insert

#CODE

table = create_object(Object, key="Blue Table", location=limbo)
chair = create_object(Object, key="Blue Chair", location=limbo)

string = "A %s and %s were created."
if DEBUG:
    table.delete()
    chair.delete()
    string += " Since debug was active, " \
         "they were deleted again."
caller.msg(string % (table, chair))

This uses Evennia’s Python API to create three objects in sequence.

Debug mode

Try to run the example script with

> @batchcode/debug tutorial_examples.example_batch_code

The batch script will run to the end and tell you it completed. You will also get messages that the button and the two pieces of furniture were created. Look around and you should see the button there. But you won’t see any chair nor a table! This is because we ran this with the /debug switch, which is directly visible as DEBUG==True inside the script. In the above example we handled this state by deleting the chair and table again.

The debug mode is intended to be used when you test out a batchscript. Maybe you are looking for bugs in your code or try to see if things behave as they should. Running the script over and over would then create an ever-growing stack of chairs and tables, all with the same name. You would have to go back and painstakingly delete them later.

Interactive mode

Interactive mode works very similar to the batch-command processor counterpart. It allows you more step-wise control over how the batch file is executed. This is useful for debugging or for picking and choosing only particular blocks to run. Use @batchcode with the /interactive flag to enter interactive mode.

> @batchcode/interactive tutorial_examples.batch_code

You should see the following:

01/02: red_button = create_object(red_button.RedButton, [...]         (hh for help)

This shows that you are on the first #CODE block, the first of only two commands in this batch file. Observe that the block has not actually been executed at this point!

To take a look at the full code snippet you are about to run, use ll (a batch-processor version of look).

from evennia.utils import create, search
from evennia.contrib.tutorial_examples import red_button
from typeclasses.objects import Object

limbo = search.objects(caller, 'Limbo', global_search=True)[0]

red_button = create.create_object(red_button.RedButton, key="Red button",
                                  location=limbo, aliases=["button"])

# caller points to the one running the script
caller.msg("A red button was created.")

Compare with the example code given earlier. Notice how the content of #HEADER has been pasted at the top of the #CODE block. Use pp to actually execute this block (this will create the button and give you a message). Use nn (next) to go to the next command. Use hh for a list of commands.

If there are tracebacks, fix them in the batch file, then use rr to reload the file. You will still be at the same code block and can rerun it easily with pp as needed. This makes for a simple debug cycle. It also allows you to rerun individual troublesome blocks - as mentioned, in a large batch file this can be very useful (don’t forget the /debug mode either).

Use nn and bb (next and back) to step through the file; e.g. nn 12 will jump 12 steps forward (without processing any blocks in between). All normal commands of Evennia should work too while working in interactive mode.

Limitations and Caveats

The batch-code processor is by far the most flexible way to build a world in Evennia. There are however some caveats you need to keep in mind.

Safety

Or rather the lack of it. There is a reason only superusers are allowed to run the batch-code processor by default. The code-processor runs without any Evennia security checks and allows full access to Python. If an untrusted party could run the code-processor they could execute arbitrary python code on your machine, which is potentially a very dangerous thing. If you want to allow other users to access the batch-code processor you should make sure to run Evennia as a separate and very limited-access user on your machine (i.e. in a ‘jail’). By comparison, the batch-command processor is much safer since the user running it is still ‘inside’ the game and can’t really do anything outside what the game commands allow them to.

No communication between code blocks

Global variables won’t work in code batch files, each block is executed as stand-alone environments. #HEADER blocks are literally pasted on top of each #CODE block so updating some header-variable in your block will not make that change available in another block. Whereas a python execution limitation, allowing this would also lead to very hard-to-debug code when using the interactive mode - this would be a classical example of “spaghetti code”.

The main practical issue with this is when building e.g. a room in one code block and later want to connect that room with a room you built in the current block. There are two ways to do this:

  • Perform a database search for the name of the room you created (since you cannot know in advance which dbref it got assigned). The problem is that a name may not be unique (you may have a lot of “A dark forest” rooms). There is an easy way to handle this though - use Tags or Aliases. You can assign any number of tags and/or aliases to any object. Make sure that one of those tags or aliases is unique to the room (like “room56”) and you will henceforth be able to always uniquely search and find it later.

  • Use the caller global property as an inter-block storage. For example, you could have a dictionary of room references in an ndb:

    #HEADER
    if caller.ndb.all_rooms is None:
        caller.ndb.all_rooms = {}
    
    #CODE
    # create and store the castle
    castle = create_object("rooms.Room", key="Castle")
    caller.ndb.all_rooms["castle"] = castle
    
    #CODE
    # in another node we want to access the castle
    castle = caller.ndb.all_rooms.get("castle")
    

    Note how we check in #HEADER if caller.ndb.all_rooms doesn’t already exist before creating the dict. Remember that #HEADER is copied in front of every #CODE block. Without that if statement we’d be wiping the dict every block!

Don’t treat a batchcode file like any Python file

Despite being a valid Python file, a batchcode file should only be run by the batchcode processor. You should not do things like define Typeclasses or Commands in them, or import them into other code. Importing a module in Python will execute base level of the module, which in the case of your average batchcode file could mean creating a lot of new objects every time.

Don’t let code rely on the batch-file’s real file path

When you import things into your batchcode file, don’t use relative imports but always import with paths starting from the root of your game directory or evennia library. Code that relies on the batch file’s “actual” location will fail. Batch code files are read as text and the strings executed. When the code runs it has no knowledge of what file those strings where once a part of.