SourceCode
Contents
- 1 the source code repository
- 2 code style guide
- 2.1 use Allman indent style.
- 2.2 use exactly one tab per indentation level
- 2.3 wrap lines longer than 100-120 characters
- 2.4 consider using indentation/alignment for function arguments
- 2.5 consider using indentation/alignment for arrays
- 2.6 make conscise use of the conditional ternary operator
- 2.7 things to consider when adding a new function
- 2.8 prefer # over // for single-line comments
- 3 examining local unsaved changes
- 4 submitting work
the source code repository
RackTables project uses git as version control system for the source code. The git repository that is used to make "tar.gz" releases is currently hosted on GitHub. It includes the following notable branches:
- master
- The main development branch, that is, the default branch for all new changes.
- maintenance-0.16.x
- maintenance-0.17.x
- maintenance-0.18.x
- maintenance-0.19.x
- Maintenance branches, that is, stable branches that serve as the base for RackTables releases.
Use the floowing commands to get a working copy of the repository:
# from the RackTables project repository git clone git://github.com/RackTables/racktables.git # or git clone https://github.com/RackTables/racktables.git # or git clone ssh://git@github.com:RackTables/racktables.git # from a GitHub fork (of your own or not) git clone git://github.com/username/racktables.git # or git clone https://github.com/username/racktables.git # or git clone ssh://git@github.com:username/racktables.git
Note that the first two schemas (git:// and https://) are always read-only and the third (ssh://) may be read-only or read-write.
code style guide
RackTables source code consists of files in several programming languages with two main cases being PHP and JavaScript. Almost all PHP source code in RackTables is original and the following style guide applies:
use Allman indent style.
Why: Allman style is very diff-friendly when it comes to control structures. In Allman style expansion of a single inner operator into a larger code block (and vice versa) leaves both the inner (controlled) and the outer (controlling) operator intact:
for ($i = 0; $i < 10; $i++) +{ doSomething ($i); + doSomethingElse(); +}
This works seamlessly at any indentation level and for all control structures (if/for/foreach/while), however deeply nested in each other. With careful indentation of case within switch it will work same well:
switch ($var) { case 1: doSomething(); break; case 2: - { - doSomethingElse(); return 3; - } case 3: + { + doSomethingCompletelyDifferent(); return 4; + } }
Being diff-friendly means being friendly to VCS (Subversion in the past, git at present). Neither the VCS nor the Allman style alone guarantee one's commits to be focused and clean. However, Allman style encourages commits that change only the lines that really need to be changed.
Exception: inline (PHP heredoc) HTML, SQL and JS code.
use exactly one tab per indentation level
Why: different people prefer different (2, 4, 8...) amount of "spaces" per indentation level. Some projects use space indentation and force everyone to hit Space (Backspace) N times each time the developer needs to indent (unindent), regardless of their own preferred value of N. RackTables makes use of Tab and enables every developer to configure their editor for their preferred "tab width" (the screenshot below stands for tab width 2). Note that the tab width must stand for the amount of spaces the editor displays, but not generates instead of, that is, in the source file tabs must remain tabs.
Exception: inline (PHP heredoc) HTML, SQL and JS code.
wrap lines longer than 100-120 characters
Why: the main motivation behind this is readability. Prior experience tells that old-school 80-columns margin just doesn't work with descriptive naming of variables and functions (i.e. descriptive == heavily wrapped). On the other hand, avoid making a readable, 200-characters long line even if it fits your monitor. Use good judgement.
consider using indentation/alignment for function arguments
Why: the purpose is twofold. On one hand it allows to pack a lot of arguments and expressions into a single function call without losing track of what is being done and no temporary variables:
usePreparedUpdateBlade ( 'Port', array ( 'name' => $port_name, 'type' => $port_type_id, 'label' => $port_label, 'reservation_comment' => $reservation_comment, 'l2address' => nullEmptyStr ($db_l2address), ), array ( 'id' => $port_id, 'object_id' => $object_id ) );
On the other hand, it allows to group function arguments add/or add comments to them:
function get8021QSyncOptions ( $vswitch, $D, // desired config $C, // cached config $R // running-config ) { $default_port = array ( 'mode' => 'access', 'allowed' => array (VLAN_DFL_ID), 'native' => VLAN_DFL_ID, ); $ret = array();
Use good judgement.
consider using indentation/alignment for arrays
Why: due to a few reasons. First, multiline array assignment/initialization can accommodate a notable amount of supplementary expressions:
$ret[$pn] = array ( 'status' => 'martian_conflict', 'left' => array_key_exists ($pn, $C) ? $C[$pn] : array ('mode' => 'none'), 'right' => array_key_exists ($pn, $R) ? $R[$pn] : array ('mode' => 'none'), );
Second, updates to arrays that serve as lists produce cleaner diffs, for example:
--- a/wwwroot/inc/remote.php +++ b/wwwroot/inc/remote.php @@ -64,6 +64,7 @@ $breedfunc = array 'jun10-get8021q-main' => 'jun10Read8021QConfig', 'jun10-xlatepushq-main' => 'jun10TranslatePushQueue', 'jun10-getallconf-main' => 'jun10SpotConfigText', + 'jun10-getlldpstatus-main' => 'jun10ReadLLDPStatus', 'ftos8-xlatepushq-main' => 'ftos8TranslatePushQueue', 'ftos8-getlldpstatus-main' => 'ftos8ReadLLDPStatus', 'ftos8-getmaclist-main' => 'ftos8ReadMacList',
Finally, it makes tree-like array initialization more transparent:
$SQLSchema = array ( 'object' => array ( 'table' => 'RackObject', 'columns' => array ( 'id' => 'id', 'name' => 'name', 'label' => 'label', 'asset_no' => 'asset_no', 'objtype_id' => 'objtype_id', 'rack_id' => '(SELECT MIN(rack_id) FROM RackSpace WHERE object_id = RackObject.id)', 'rack_id_2' => "(SELECT MIN(parent_entity_id) FROM EntityLink WHERE child_entity_type='object' AND child_entity_id = RackObject.id AND parent_entity_type = 'rack')", 'container_id' => "(SELECT MIN(parent_entity_id) FROM EntityLink WHERE child_entity_type='object' AND child_entity_id = RackObject.id AND parent_entity_type = 'object')", 'container_name' => '(SELECT name FROM RackObject WHERE id = container_id)', 'container_objtype_id' => '(SELECT objtype_id FROM RackObject WHERE id = container_id)', 'has_problems' => 'has_problems', 'comment' => 'comment', 'nports' => '(SELECT COUNT(*) FROM Port WHERE object_id = RackObject.id)', '8021q_domain_id' => '(SELECT domain_id FROM VLANSwitch WHERE object_id = id LIMIT 1)', '8021q_template_id' => '(SELECT template_id FROM VLANSwitch WHERE object_id = id LIMIT 1)', ), 'keycolumn' => 'id', 'ordcolumns' => array ('RackObject.name'), ),
make conscise use of the conditional ternary operator
Why: to use the language element in making the code look closer to what the code is intended to do. First, don't reinvent the conditional ternary operator with if like this:
if ($row !== FALSE) return $row[0]; else return NULL;
That block of code returns a value in either case, exactly one return would do the job:
return $row !== FALSE ? $row[0] : NULL;
Second, don't nest ternary conditionals, it is hard to read:
$ret = ($ret > 0 ? 1 : ($ret < 0 ? -1 : 0));
When a standalone ternary conditional isn't sufficient, use temporary variables or if's.
things to consider when adding a new function
- First, is it possible to use one of the existing functions or a combination of these?
- Carefully decide which file the function should be in (depends on its purpose and relation to other functions).
- Consider headlessCamelCase for the function name.
- Try to keep functions focused, that is, small, doing just one thing but doing it very well.
prefer # over // for single-line comments
Why: it saves keystokes. However, when adding a comment to a code that already uses //, be consistent with what is already there.
examining local unsaved changes
git status git diff
submitting work
In the case of a GitHub fork of RackTables publishing changes is as simple as using git push. To merge your changes into the upstream RackTables notify the dev team by any convenient mean (GitGub pull request or email).
In the case of a private repository use a git format-patch command to produce a patch and send it to the developers.