CONTRIBUTING

Reporting Issues

If you ever encounter a bug in-game, the best way to let a coder know about it is with our GitHub Issue Tracker. Please make sure you use the supplied issue template, and include the round ID for the server.

(If you don't have an account, making a new one takes only one minute.)

If you have difficulty, ask for help in the #coding-general channel on our discord.

Introduction

Hello and welcome to /tg/station's contributing page. You are here because you are curious or interested in contributing - thank you! Everyone is free to contribute to this project as long as they follow the simple guidelines and specifications below; at /tg/station, we strive to maintain code stability and maintainability, and to do that, we need all pull requests to hold up to those specifications. It's in everyone's best interests - including yours! - if the same bug doesn't have to be fixed twice because of duplicated code.

First things first, we want to make it clear how you can contribute (if you've never contributed before), as well as the kinds of powers the team has over your additions, to avoid any unpleasant surprises if your pull request is closed for a reason you didn't foresee.

Getting Started

/tg/station doesn't have a list of goals and features to add; we instead allow freedom for contributors to suggest and create their ideas for the game. That doesn't mean we aren't determined to squash bugs, which unfortunately pop up a lot due to the deep complexity of the game. Here are some useful starting guides, if you want to contribute or if you want to know what challenges you can tackle with zero knowledge about the game's code structure.

If you want to contribute the first thing you'll need to do is set up Git so you can download the source code. After setting it up, optionally navigate your git commandline to the project folder and run the command: 'git config blame.ignoreRevsFile .git-blame-ignore-revs'

We have a list of guides on the wiki that will help you get started contributing to /tg/station with Git and Dream Maker. For beginners, it is recommended you work on small projects like bugfixes at first. If you need help learning to program in BYOND, check out this repository of resources.

There is an open list of approachable issues for your inspiration here.

You can of course, as always, ask for help on the discord channels, or the forums, We're just here to have fun and help out, so please don't expect professional support.

Meet the Team

Headcoder

The Headcoder is responsible for controlling, adding, and removing maintainers from the project. In addition to filling the role of a normal maintainer, they have sole authority on who becomes a maintainer, as well as who remains a maintainer and who does not.

Maintainers

Maintainers are quality control. If a proposed pull request doesn't meet the following specifications, they can request you to change it, or simply just close the pull request. Maintainers are required to give a reason for closing the pull request.

Maintainers can revert your changes if they feel they are not worth maintaining or if they did not live up to the quality specifications.

Specifications

As mentioned before, you are expected to follow these specifications in order to make everyone's lives easier. It'll save both your time and ours, by making sure you don't have to make any changes and we don't have to ask you to. Thank you for reading this section!

Object Oriented Code

As BYOND's Dream Maker (henceforth "DM") is an object-oriented language, code must be object-oriented when possible in order to be more flexible when adding content to it. If you don't know what "object-oriented" means, we highly recommend you do some light research to grasp the basics.

All BYOND paths must contain the full path

(i.e. absolute pathing)

DM will allow you nest almost any type keyword into a block, such as:

datum
	datum1
		var
			varname1 = 1
			varname2
			static
				varname3
				varname4
		proc
			proc1()
				code
			proc2()
				code

		datum2
			varname1 = 0
			proc
				proc3()
					code
			proc2()
				. = ..()
				code

The use of this is not allowed in this project as it makes finding definitions via full text searching next to impossible. The only exception is the variables of an object may be nested to the object, but must not nest further.

The previous code made compliant:

/datum/datum1
	var/varname1
	var/varname2
	var/static/varname3
	var/static/varname4

/datum/datum1/proc/proc1()
	code
/datum/datum1/proc/proc2()
	code
/datum/datum1/datum2
	varname1 = 0
/datum/datum1/datum2/proc/proc3()
	code
/datum/datum1/datum2/proc2()
	. = ..()
	code

No overriding type safety checks

The use of the : operator to override type safety checks is not allowed. You must cast the variable to the proper type.

Type paths must begin with a /

eg: /datum/thing, not datum/thing

Type paths must be snake case

eg: /datum/thing/blue_bird, not datum/thing/BLUEBIRD or datum/thing/BlueBird or datum/thing/Bluebird or datum/thing/blueBird

Datum type paths must began with "datum"

In DM, this is optional, but omitting it makes finding definitions harder.

Do not use text/string based type paths

It is rarely allowed to put type paths in a text format, as there are no compile errors if the type path no longer exists. Here is an example:

//Good
var/path_type = /obj/item/baseball_bat

//Bad
var/path_type = "/obj/item/baseball_bat"

Use var/name format when declaring variables

While DM allows other ways of declaring variables, this one should be used for consistency.

Tabs, not spaces

You must use tabs to indent your code, NOT SPACES.

(You may use spaces to align something, but you should tab to the block level first, then add the remaining spaces)

No hacky code

Hacky code, such as adding specific checks, is highly discouraged and only allowed when there is no other option. (Protip: 'I couldn't immediately think of a proper way so thus there must be no other option' is not gonna cut it here! If you can't think of anything else, say that outright and admit that you need help with it. Maintainers exist for exactly that reason.)

You can avoid hacky code by using object-oriented methodologies, such as overriding a function (called "procs" in DM) or sectioning code into functions and then overriding them as required.

No duplicated code

Copying code from one place to another may be suitable for small, short-time projects, but /tg/station is a long-term project and highly discourages this.

Instead you can use object orientation, or simply placing repeated code in a function, to obey this specification easily.

Startup/Runtime tradeoffs with lists and the "hidden" init proc

First, read the comments in this BYOND thread, starting where the link takes you.

There are two key points here:

Defining a list in the variable's definition calls a hidden proc - init. If you have to define a list at startup, do so in New() (or preferably Initialize()) and avoid the overhead of a second call (Init() and then New())
It also consumes more memory to the point where the list is actually required, even if the object in question may never use it!

Remember: although this tradeoff makes sense in many cases, it doesn't cover them all. Think carefully about your addition before deciding if you need to use it.

Prefer `Initialize()` over `New()` for atoms

Our game controller is pretty good at handling long operations and lag, but it can't control what happens when the map is loaded, which calls New for all atoms on the map. If you're creating a new atom, use the Initialize proc to do what you would normally do in New. This cuts down on the number of proc calls needed when the world is loaded. See here for details on Initialize: https://github.com/tgstation/tgstation/blob/34775d42a2db4e0f6734560baadcfcf5f5540910/code/game/atoms.dm#L166 While we normally encourage (and in some cases, even require) bringing out of date code up to date when you make unrelated changes near the out of date code, that is not the case for New -> Initialize conversions. These systems are generally more dependant on parent and children procs so unrelated random conversions of existing things can cause bugs that take months to figure out.

No magic numbers or strings

This means stuff like having a "mode" variable for an object set to "1" or "2" with no clear indicator of what that means. Make these #defines with a name that more clearly states what it's for. For instance:

/datum/proc/do_the_thing(thing_to_do)
	switch(thing_to_do)
		if(1)
			(...)
		if(2)
			(...)

There's no indication of what "1" and "2" mean! Instead, you'd do something like this:

#define DO_THE_THING_REALLY_HARD 1
#define DO_THE_THING_EFFICIENTLY 2
/datum/proc/do_the_thing(thing_to_do)
	switch(thing_to_do)
		if(DO_THE_THING_REALLY_HARD)
			(...)
		if(DO_THE_THING_EFFICIENTLY)
			(...)

This is clearer and enhances readability of your code! Get used to doing it!

Control statements

(if, while, for, etc)

No control statement may contain code on the same line as the statement (if (blah) return)
All control statements comparing a variable to a number should use the formula of thing operator number, not the reverse (eg: if (count <= 10) not if (10 >= count))

Use early return

Do not enclose a proc in an if-block when returning on a condition is more feasible This is bad:

/datum/datum1/proc/proc1()
	if (thing1)
		if (!thing2)
			if (thing3 == 30)
				do stuff

This is good:

/datum/datum1/proc/proc1()
	if (!thing1)
		return
	if (thing2)
		return
	if (thing3 != 30)
		return
	do stuff

This prevents nesting levels from getting deeper then they need to be.

Use our time defines

The codebase contains some defines which will automatically multiply a number by the correct amount to get a number in deciseconds. Using these is preffered over using a literal amount in deciseconds.

The defines are as follows:

SECONDS
MINUTES
HOURS

This is bad:

/datum/datum1/proc/proc1()
	if(do_after(mob, 15))
		mob.dothing()

This is good:

/datum/datum1/proc/proc1()
	if(do_after(mob, 1.5 SECONDS))
		mob.dothing()

Getters and setters

Avoid getter procs. They are useful tools in languages with that properly enforce variable privacy and encapsulation, but DM is not one of them. The upfront cost in proc overhead is met with no benefits, and it may tempt to develop worse code.

This is bad:

/datum/datum1/proc/simple_getter()
	return gotten_variable

Prefer to either access the variable directly or use a macro/define.

Make usage of variables or traits, set up through condition setters, for a more maintainable alternative to compex and redefined getters.

These are bad:

/datum/datum1/proc/complex_getter()
	return condition ? VALUE_A : VALUE_B

/datum/datum1/child_datum/complex_getter()
	return condition ? VALUE_C : VALUE_D

This is good:

/datum/datum1
	var/getter_turned_into_variable

/datum/datum1/proc/set_condition(new_value)
	if(condition == new_value)
		return
	condition = new_value
	on_condition_change()

/datum/datum1/proc/on_condition_change()
	getter_turned_into_variable = condition ? VALUE_A : VALUE_B

/datum/datum1/child_datum/on_condition_change()
	getter_turned_into_variable = condition ? VALUE_C : VALUE_D

Develop Secure Code

Player input must always be escaped safely, we recommend you use stripped_input in all cases where you would use input. Essentially, just always treat input from players as inherently malicious and design with that use case in mind
Calls to the database must be escaped properly - use sanitizeSQL to escape text based database entries from players or admins, and isnum() for number based database entries from players or admins.
All calls to topics must be checked for correctness. Topic href calls can be easily faked by clients, so you should ensure that the call is valid for the state the item is in. Do not rely on the UI code to provide only valid topic calls, because it won't.
Information that players could use to metagame (that is, to identify round information and/or antagonist type via information that would not be available to them in character) should be kept as administrator only.
It is recommended as well you do not expose information about the players - even something as simple as the number of people who have readied up at the start of the round can and has been used to try to identify the round type.
Where you have code that can cause large-scale modification and FUN, make sure you start it out locked behind one of the default admin roles - use common sense to determine which role fits the level of damage a function could do.

Files

Because runtime errors do not give the full path, try to avoid having files with the same name across folders.
File names should not be mixed case, or contain spaces or any character that would require escaping in a uri.
Files and path accessed and referenced by code above simply being #included should be strictly lowercase to avoid issues on filesystems where case matters.

SQL

Do not use the shorthand sql insert format (where no column names are specified) because it unnecessarily breaks all queries on minor column changes and prevents using these tables for tracking outside related info such as in a connected site/forum.
All changes to the database's layout(schema) must be specified in the database changelog in SQL, as well as reflected in the schema files
Any time the schema is changed the schema_revision table and DB_MAJOR_VERSION or DB_MINOR_VERSION defines must be incremented.
Queries must never specify the database, be it in code, or in text files in the repo.
Primary keys are inherently immutable and you must never do anything to change the primary key of a row or entity. This includes preserving auto increment numbers of rows when copying data to a table in a conversion script. No amount of bitching about gaps in ids or out of order ids will save you from this policy.
The ttl for data from the database is 10 seconds. You must have a compelling reason to store and reuse data for longer then this.
Do not write stored and transformed data to the database, instead, apply the transformation to the data in the database directly.
- ie: SELECTing a number from the database, doubling it, then updating the database with the doubled number. If the data in the database changed between step 1 and 3, you'll get an incorrect result. Instead, directly double it in the update query. UPDATE table SET num = num*2 instead of UPDATE table SET num = [num].
- if the transformation is user provided (such as allowing a user to edit a string), you should confirm the value being updated did not change in the database in the intervening time before writing the new user provided data by checking the old value with the current value in the database, and if it has changed, allow the user to decide what to do next.

Mapping Standards

TGM Format & Map Merge
- All new maps submitted to the repo through a pull request must be in TGM format (unless there is a valid reason present to have it in the default BYOND format.) This is done using the Map Merge utility included in the repo to convert the file to TGM format.
- Likewise, you MUST run Map Merge prior to opening your PR when updating existing maps to minimize the change differences (even when using third party mapping programs such as FastDMM.)
  - Failure to run Map Merge on a map after using third party mapping programs (such as FastDMM) greatly increases the risk of the map's key dictionary becoming corrupted by future edits after running map merge. Resolving the corruption issue involves rebuilding the map's key dictionary; id est rewriting all the keys contained within the map by reconverting it from BYOND to TGM format - which creates very large differences that ultimately delay the PR process and is extremely likely to cause merge conflicts with other pull requests.
Variable Editing (Var-edits)
- While var-editing an item within the editor is perfectly fine, it is preferred that when you are changing the base behavior of an item (how it functions) that you make a new subtype of that item within the code, especially if you plan to use the item in multiple locations on the same map, or across multiple maps. This makes it easier to make corrections as needed to all instances of the item at one time as opposed to having to find each instance of it and change them all individually.
  - Subtypes only intended to be used on away mission or ruin maps should be contained within a .dm file with a name corresponding to that map within code\modules\awaymissions or code\modules\ruins respectively. This is so in the event that the map is removed, that subtype will be removed at the same time as well to minimize leftover/unused data within the repo.
- Please attempt to clean out any dirty variables that may be contained within items you alter through var-editing. For example, due to how DM functions, changing the pixel_x variable from 23 to 0 will leave a dirty record in the map's code of pixel_x = 0. Likewise this can happen when changing an item's icon to something else and then back. This can lead to some issues where an item's icon has changed within the code, but becomes broken on the map due to it still attempting to use the old entry.
- Areas should not be var-edited on a map to change it's name or attributes. All areas of a single type and it's altered instances are considered the same area within the code, and editing their variables on a map can lead to issues with powernets and event subsystems which are difficult to debug.

User Interfaces

All new player-facing user interfaces must use TGUI.
Raw HTML is permitted for admin and debug UIs.
Documentation for TGUI can be found at:
- tgui/README.md
- tgui/tutorial-and-examples.md

Other Notes

Code should be modular where possible; if you are working on a new addition, then strongly consider putting it in its own file unless it makes sense to put it with similar ones (i.e. a new tool would go in the "tools.dm" file)
Bloated code may be necessary to add a certain feature, which means there has to be a judgement over whether the feature is worth having or not. You can help make this decision easier by making sure your code is modular.
You are expected to help maintain the code that you add, meaning that if there is a problem then you are likely to be approached in order to fix any issues, runtimes, or bugs.
Do not divide when you can easily convert it to multiplication. (ie 4/2 should be done as 4*0.5)
If you used regex to replace code during development of your code, post the regex in your PR for the benefit of future developers and downstream users.
Changes to the /config tree must be made in a way that allows for updating server deployments while preserving previous behaviour. This is due to the fact that the config tree is to be considered owned by the user and not necessarily updated alongside the remainder of the code. The code to preserve previous behaviour may be removed at some point in the future given the OK by maintainers.
The dlls section of tgs3.json is not designed for dlls that are purely call()()ed since those handles are closed between world reboots. Only put in dlls that may have to exist between world reboots.

Enforced not enforced

The following coding styles are not only not enforced at all, but are generally frowned upon to change for little to no reason:

English/British spelling on var/proc names
- Color/Colour - both are fine, but keep in mind that BYOND uses color as a base variable
Spaces after control statements
- if() and if () - nobody cares!

Operators

Spacing

(this is not strictly enforced, but more a guideline for readability's sake)

Operators that should be separated by spaces
- Boolean and logic operators like &&, || <, >, ==, etc (but not !)
- Bitwise AND &
- Argument separator operators like , (and ; when used in a forloop)
- Assignment operators like = or += or the like
Operators that should not be separated by spaces
- Bitwise OR |
- Access operators like . and :
- Parentheses ()
- logical not !

Math operators like +, -, /, *, etc are up in the air, just choose which version looks more readable.

Use

Bitwise AND - '&'
- Should be written as bitfield & bitflag NEVER bitflag & bitfield, both are valid, but the latter is confusing and nonstandard.
Associated lists declarations must have their key value quoted if it's a string
- WRONG: list(a = "b")
- RIGHT: list("a" = "b")

Dream Maker Quirks/Tricks

Like all languages, Dream Maker has its quirks, some of them are beneficial to us, like these

In-To for-loops

for(var/i = 1, i <= some_value, i++) is a fairly standard way to write an incremental for loop in most languages (especially those in the C family), but DM's for(var/i in 1 to some_value) syntax is oddly faster than its implementation of the former syntax; where possible, it's advised to use DM's syntax. (Note, the to keyword is inclusive, so it automatically defaults to replacing <=; if you want < then you should write it as 1 to some_value-1).

HOWEVER, if either some_value or i changes within the body of the for (underneath the for(...) header) or if you are looping over a list AND changing the length of the list then you can NOT use this type of for-loop!

for(var/A in list) VS for(var/i in 1 to list.len)

The former is faster than the latter, as shown by the following profile results: https://file.house/zy7H.png Code used for the test in a readable format: https://pastebin.com/w50uERkG

Istypeless for loops

A name for a differing syntax for writing for-each style loops in DM. It's NOT DM's standard syntax, hence why this is considered a quirk. Take a look at this:

var/list/bag_of_items = list(sword, apple, coinpouch, sword, sword)
var/obj/item/sword/best_sword
for(var/obj/item/sword/S in bag_of_items)
	if(!best_sword || S.damage > best_sword.damage)
		best_sword = S

The above is a simple proc for checking all swords in a container and returning the one with the highest damage, and it uses DM's standard syntax for a for-loop by specifying a type in the variable of the for's header that DM interprets as a type to filter by. It performs this filter using istype() (or some internal-magic similar to istype() - this is BYOND, after all). This is fine in its current state for bag_of_items, but if bag_of_items contained ONLY swords, or only SUBTYPES of swords, then the above is inefficient. For example:

var/list/bag_of_swords = list(sword, sword, sword, sword)
var/obj/item/sword/best_sword
for(var/obj/item/sword/S in bag_of_swords)
	if(!best_sword || S.damage > best_sword.damage)
		best_sword = S

specifies a type for DM to filter by.

With the previous example that's perfectly fine, we only want swords, but here the bag only contains swords? Is DM still going to try to filter because we gave it a type to filter by? YES, and here comes the inefficiency. Wherever a list (or other container, such as an atom (in which case you're technically accessing their special contents list, but that's irrelevant)) contains datums of the same datatype or subtypes of the datatype you require for your loop's body, you can circumvent DM's filtering and automatic istype() checks by writing the loop as such:

var/list/bag_of_swords = list(sword, sword, sword, sword)
var/obj/item/sword/best_sword
for(var/s in bag_of_swords)
	var/obj/item/sword/S = s
	if(!best_sword || S.damage > best_sword.damage)
		best_sword = S

Of course, if the list contains data of a mixed type then the above optimisation is DANGEROUS, as it will blindly typecast all data in the list as the specified type, even if it isn't really that type, causing runtime errors.

Dot variable

Like other languages in the C family, DM has a . or "Dot" operator, used for accessing variables/members/functions of an object instance. eg:

var/mob/living/carbon/human/H = YOU_THE_READER
H.gib()

However, DM also has a dot variable, accessed just as . on its own, defaulting to a value of null. Now, what's special about the dot operator is that it is automatically returned (as in the return statement) at the end of a proc, provided the proc does not already manually return (return count for example.) Why is this special?

With . being everpresent in every proc, can we use it as a temporary variable? Of course we can! However, the . operator cannot replace a typecasted variable - it can hold data any other var in DM can, it just can't be accessed as one, although the . operator is compatible with a few operators that look weird but work perfectly fine, such as: .++ for incrementing .'s value, or .[1] for accessing the first element of ., provided that it's a list.

Globals versus static

DM has a var keyword, called global. This var keyword is for vars inside of types. For instance:

mob
	var
		global
			thing = TRUE

This does NOT mean that you can access it everywhere like a global var. Instead, it means that that var will only exist once for all instances of its type, in this case that var will only exist once for all mobs - it's shared across everything in its type. (Much more like the keyword static in other languages like PHP/C++/C#/Java)

Isn't that confusing?

There is also an undocumented keyword called static that has the same behaviour as global but more correctly describes BYOND's behaviour. Therefore, we always use static instead of global where we need it, as it reduces suprise when reading BYOND code.

Pull Request Process

There is no strict process when it comes to merging pull requests. Pull requests will sometimes take a while before they are looked at by a maintainer; the bigger the change, the more time it will take before they are accepted into the code. Every team member is a volunteer who is giving up their own time to help maintain and contribute, so please be courteous and respectful. Here are some helpful ways to make it easier for you and for the maintainers when making a pull request.

Make sure your pull request complies to the requirements outlined here
You are going to be expected to document all your changes in the pull request. Failing to do so will mean delaying it as we will have to question why you made the change. On the other hand, you can speed up the process by making the pull request readable and easy to understand, with diagrams or before/after data. Should you be optimizing a routine you must provide proof by way of profiling that your changes are faster.
We ask that you use the changelog system to document your change, which prevents our players from being caught unaware by changes - you can find more information about this on this wiki page.
If you are proposing multiple changes, which change many different aspects of the code, you are expected to section them off into different pull requests in order to make it easier to review them and to deny/accept the changes that are deemed acceptable.
If your pull request is accepted, the code you add no longer belongs exclusively to you but to everyone; everyone is free to work on it, but you are also free to support or object to any changes being made, which will likely hold more weight, as you're the one who added the feature. It is a shame this has to be explicitly said, but there have been cases where this would've saved some trouble.
Please explain why you are submitting the pull request, and how you think your change will be beneficial to the game. Failure to do so will be grounds for rejecting the PR.
If your pull request is not finished make sure it is at least testable in a live environment. Pull requests that do not at least meet this requirement will be closed. You may request a maintainer reopen the pull request when you're ready, or make a new one.
While we have no issue helping contributors (and especially new contributors) bring reasonably sized contributions up to standards via the pull request review process, larger contributions are expected to pass a higher bar of completeness and code quality before you open a pull request. Maintainers may close such pull requests that are deemed to be substantially flawed. You should take some time to discuss with maintainers or other contributors on how to improve the changes.

Porting features/sprites/sounds/tools from other codebases

If you are porting features/tools from other codebases, you must give them credit where it's due. Typically, crediting them in your pull request and the changelog is the recommended way of doing it. Take note of what license they use though, porting stuff from AGPLv3 and GPLv3 codebases are allowed.

Regarding sprites & sounds, you must credit the artist and possibly the codebase. All /tg/station assets including icons and sound are under a Creative Commons 3.0 BY-SA license unless otherwise indicated. However if you are porting assets from GoonStation or usually any assets under the Creative Commons 3.0 BY-NC-SA license are to go into the 'goon' folder of the /tg/station codebase.

Banned content

Do not add any of the following in a Pull Request or risk getting the PR closed:

National Socialist Party of Germany content, National Socialist Party of Germany related content, or National Socialist Party of Germany references
Code where one line of code is split across mutiple lines (except for multiple, separate strings and comments; in those cases, existing longer lines must not be split up)
Code adding, removing, or updating the availability of alien races/species/human mutants without prior approval. Pull requests attempting to add or remove features from said races/species/mutants require prior approval as well.
Code which violates GitHub's terms of service.

Just because something isn't on this list doesn't mean that it's acceptable. Use common sense above all else.

A word on Git

This repository uses LF line endings for all code as specified in the .gitattributes and .editorconfig files.

Unless overridden or a non standard git binary is used the line ending settings should be applied to your clone automatically.

Note: VSC requires an extension to take advantage of editorconfig.

Files

CONTRIBUTING.md

Latest commit

History

CONTRIBUTING.md

File metadata and controls

CONTRIBUTING

Reporting Issues

Introduction

Getting Started

Meet the Team

Specifications

Object Oriented Code

All BYOND paths must contain the full path

No overriding type safety checks

Type paths must begin with a /

Type paths must be snake case

Datum type paths must began with "datum"

Do not use text/string based type paths

Use var/name format when declaring variables

Tabs, not spaces

No hacky code

No duplicated code

Startup/Runtime tradeoffs with lists and the "hidden" init proc

Prefer Initialize() over New() for atoms

No magic numbers or strings

Control statements

Use early return

Use our time defines

Getters and setters

Develop Secure Code

Files

SQL

Mapping Standards

User Interfaces

Other Notes

Enforced not enforced

Operators

Spacing

Use

Dream Maker Quirks/Tricks

In-To for-loops

for(var/A in list) VS for(var/i in 1 to list.len)

Istypeless for loops

Dot variable

Globals versus static

Pull Request Process

Porting features/sprites/sounds/tools from other codebases

Banned content

A word on Git

Prefer `Initialize()` over `New()` for atoms