Website & Domain Tools

The Domain Lab

Website › Configuration › Domain Lab. Paste any domain, pick any model, press Run it. You get three answers:

• The count — how many records the domain really matches.

• A sample — the first few of them, by name, so "1,204 records" stops being an abstraction.

• Leaf by leaf — every condition with two numbers next to it: what it matches alone, and the running total with everything before it applied. The lonely count is how you spot the condition that matches nothing by itself — the typo'd value, the wrong operator. The running one is how you spot the condition that undoes everything before it.

Everything is counted with your rights: the ACLs and the record rules that apply to you apply to the lab. A count you cannot reproduce in a list view would teach you the wrong thing, so the lab never shows you one.

Errors come back as answers, not tracebacks. A domain lab where a typo raises a 500 is a lab nobody pastes drafts into.

The eval context carries the scalars record rules and filters use — uid, company_ids, company_id, context_today(), datetime, relativedelta — so the domain you test is the domain you ship. It deliberately does not put a user record in scope: this is a free-text box any internal user can type into, and a live recordset there would be a way to read and write the whole database through it. Write uid, not user.id.

Snippets. A domain somebody got right once is worth keeping. Name it, save it, and the next person loads it instead of retyping it slightly differently into their fourth record rule. Snippets are readable by every internal user and written by website designers.

Alias domains

Each website carries a list of alias hostnames next to its canonical domain — one per line, on the website form. A request arriving on any alias is routed to that website exactly as if it had used the real domain: shop.example.com and www.example.co.uk reach the same site without anybody editing nginx.

Schemes, ports and case do not confuse it — hostnames arrive however the proxy felt like writing them. An alias cannot collide with another website's real domain or another alias: the form refuses it, so one site can never silently hijack another's traffic. The alias map is cached and rebuilt only when a website changes, so the routing costs a dictionary lookup, not a query, on every request.

Copying a website's configuration

The second website always starts empty. Website › Configuration › Copy Website Configuration starts it right instead: pick a source, a target, and what to carry over —

• The menu tree, with its shape — mega-menu dropdowns and all. Links are kept as URLs; the pages behind them are not cloned. If the new website has a page at that path the menu just works, and if it does not, the menu is honest about what is missing — which beats forty silently duplicated pages and the duplicate-content penalty that comes with them.

• The social links and the logo. A deliberate whitelist, not a pattern match — a field a future version adds is copied only after somebody decides it should be.

• The homepage setting.

The wizard says exactly what it copied, and copying a website onto itself is refused.

What it deliberately does not do

• It does not clone pages, products or any content. Configuration only.

• It does not copy configuration across companies — a website of one company onto a website of another is refused, because it is almost never what is meant.

• It does not validate a snippet's domain at save time — the lab is where domains are tested, and a draft is allowed to be wrong there.

• The lab is an internal tool: portal and public users are refused at the server, not just hidden from the menu.