Skip to content

Table columns

Every list in Danbyte lets you choose which columns to show and in what order. Your layout is remembered per table, and an administrator can publish a shared default for everyone.

Customizing a table

Open the Columns menu in the table's toolbar. It has two parts:

  • Shown - the columns on screen, in table order. Drag a column by its grip to move it; untick it to hide it.
  • Available - everything else the table can show, in sections, each sorted alphabetically. Tick one and it joins the end of Shown.

Press Save to keep the layout. Closing the menu without saving discards your changes. Reset (when you have a saved layout) goes back to the tenant default. A long list gets a search box at the top; while you type, dragging is off.

Note

A few columns (like the row-select checkbox and the row-actions menu) always stay in place and can't be moved or hidden.

Pinned columns

Click the pin beside a column under Shown to pin it, then Save. A pinned column stays in its place in the table; it only sticks to the left edge once you scroll sideways past it, and scrolling back sets it down in its place again. Pin Name and Coverage on Devices: both sit where they always do, Name sticks with the first scroll, and Coverage sticks next to it once you scroll past it. Several pins stack up at the left in table order. On a table with checkboxes the checkbox column sticks too whenever anything is pinned. A thin line marks the edge of the stuck columns while they are stuck. Click the pin again to unpin. Tables on detail tabs have the same pin in their Columns menu; a table that keeps no saved layout keeps its pins until you leave the page.

  • Pins are saved with the rest of your layout, per table, and a published tenant default carries them too.
  • Stuck columns take at most 60% of the table's width. A pin that would go past that scrolls like any other column, so there is always something left to scroll.
  • On a phone-width window (a table under 640 px wide) pins are ignored.
  • Pinning never moves a column. Hiding a column drops its pin.

Wide tables

A table with more columns than fit the window scrolls sideways. Its horizontal scrollbar floats at the bottom of the window while the table is on screen, so you can scroll sideways without first scrolling to the last row, and a soft shadow on the left or right edge shows there are more columns that way. Shift + mouse wheel and sideways trackpad scrolling work too.

Every field is a column

A table is not limited to the columns its page was designed with. The Available part lists:

Section What is in it
Columns The table's own columns you have hidden.
Fields Every other field the list's rows carry - a device's asset tag, airflow, rack position, created date.
Related Linked objects, including one step further out: a device's Region (its site's region), location, rack, platform, primary / secondary / OOB IP, config template. An interface's Link peer reads as the far device and port (sw1 · Gi1/0/1).
Custom fields Every custom field defined for that object type.

These columns start hidden, so a table looks the same until you tick one. They sort when you click their header (numbers by value, text in natural order, empty cells last), and they are included in exports.

The list is built from what the list's rows actually contain (see GET /api/list-fields/ in the API reference), so a field added to a list later appears here on its own - nobody maintains a list of columns per table. It never adds data: you only see fields the list already sends to you. Device fields an administrator has switched off (Settings → Device fields, e.g. airflow and cluster) are not offered, and neither are fields with nothing to show in a cell - internal flags, sets of figures, and counts the detail page computes for one object only.

Filtering by one of these columns

The filter rail keeps its designed facets. To filter on any other field, use the advanced filter expression (for example location.name = "Hall A", site.region = "Nordics" or custom_fields.owner = "noc").

Filter rail

The filter rail beside a list has one group per facet - status, role, manufacturer, tags - each counting the rows in every value.

  • Long facets. A facet with more than 10 values gets a search box and shows its 8 most common values; Show all lists the rest. A ticked value always stays in sight.
  • Hiding a facet. Hover a facet's name and click the eye to hide it on that list - Advanced included. Hidden facets are listed under Hidden filters at the foot of the rail; click one to bring it back. A hidden facet with a selection keeps showing, so no filter applies out of sight.

Hidden facets are saved per list for you, like a column layout (stored as the facets-<list> layout, e.g. facets-devices), and do not affect anyone else.

A list's filters live in its address, so a reload, Back from an object, a bookmark or a pasted link all show the same filtered list:

/devices?status=active,planned&role=leaf&tag=core&q=sw-01
  • One parameter per facet, named after it: status, role, site, type, … The Tags facet is tag. Values are ids and slugs, not display names.
  • Several values are comma-separated and match any of them. A value that contains a comma is written with a backslash before it (a\,b).
  • Ranges read min..max, either side optional: util=80...
  • The search box is q, written once typing pauses. The advanced filter is expr.
  • Nothing selected means no parameter. A value the list does not have (a deleted tag, a typo) is ignored rather than emptying the list.

Changing a filter replaces the current history entry instead of adding one, so Back leaves the list instead of undoing filters one by one.

Dashboard and cross-object links (Devices by status, a rack's port utilisation, a certificate-health tile) use the same parameters, so the filter they open with shows in the rail and can be cleared there. A saved view writes its filters into the address when applied.

Tags

  • In a list, click a tag chip to filter the list by that tag; click it again to clear it. Active tags get a ring, and several tags match any of them.
  • All objects with a tag: beside a ticked tag in the Tags facet, the arrow opens the tag's Tagged objects tab, across every object type.
  • On an object's page, a tag links to its list filtered by that tag, for example a device's core tag opens /devices?tag=core.

Searching a table

A list page searches with the box in its header. A table on a detail tab - a rack's devices, a device's interfaces, a site's prefixes, a virtualization source's VMs, an object's change log - has its own search box above the rows when it can run long. It matches the words a row shows in its shown columns, ignoring case: names, addresses, statuses, related objects by name. Hidden columns are not searched.

The box filters the rows the tab has loaded. A tab that loads only the first part of a long list says so under the table - Searching the first 500 of 1,234 - and the full list page searches the rest.

Very long lists

A list page loads up to 10,000 rows, 1,000 a request, and 1,000 in all for IP addresses. When the tenant holds more, the page says so:

  • the count beside the title reads 1,000 of 40,960, in amber;
  • a notice over the table reads Showing the first 1,000 of 40,960 - search to narrow the list;
  • the filter rail notes that its counts cover the loaded rows only;
  • Download names the rows it holds - First 1,000 of 40,960 rows - in its menu and on the HTML and print page.

The header search runs on the server, so it reaches every row; the filter rail and the advanced filter refine the rows that arrived. This applies to IP addresses, interfaces, cables, devices, prefixes, VLANs and virtual machines.

Sorting

Click a column header to sort by it; click it again to reverse. Text sorts in natural order, the way you count: DIMM 1, DIMM 2 … DIMM 10, DIMM 11, and Ethernet1/2 before Ethernet1/10, with case ignored. Numbers and dates sort by value. The order is the same whatever language your browser uses. Before you click anything, a list shows rows in the order the server sends them, which is the same natural order by name.

Selecting rows

On a list with checkboxes, the header checkbox ticks the rows on the current page. When the list runs to more pages, Select all N appears beside the count above the table and ticks every row the filters show; Clear undoes it. A tick belongs to the object, not to its row position, so sorting, filtering or a refresh never moves it onto another row. A row that leaves the list - filtered out, hidden by the table's own search box, or deleted - leaves the selection too, and stays unticked when it comes back, so a bulk action never reaches a row you can't find. Ticked rows on other pages stay ticked; that is what Select all N is for.

Large selections

One bulk call takes at most 1000 rows (2000 on the MAC list), so a bigger selection goes in batches of that size, one after another. Where the confirmation first counts what would happen - what a delete removes, releases and keeps, or what a MAC removal reaches - it adds the batches up into one answer. While the action runs, its button shows the batch on its way: Deleting… 2 / 3.

If a batch fails, the run stops there. The batches before it stand, the ones after it are never sent, and the message says how many rows went through and why the rest did not. What is left stays selected, so pressing the button again finishes the job - a delete, an edit or an add is safe to repeat.

Rename and Clone on a component table check every new name against the whole selection, so they take at most 1000 rows at a time and say so before anything runs.

Where your settings live

Manage all your saved table layouts in one place under User → Preferences, where each table shows its current state with a Reset option. Tabs on detail pages remember their layout too, but are left out of that list.

Shared defaults (administrators)

If you can manage users, you can also publish a single table's starting layout for your whole tenant under Settings → Table layouts → Tenant defaults. These defaults from before presets keep working: they apply after every preset (and a locked one still wins over everything).

Action Effect
Publish Makes your current layout of that table the tenant default.
Lock Forces that default - everyone uses it and can't change their own.
Unlock Lets people customize again, starting from the default.
Clear Removes the tenant default entirely.

Layout presets (administrators)

A layout preset is a named set of table layouts - columns, order, hidden and pinned - that you share as the default for other people. Make one under Settings → Table layouts → Layout presets with New preset from my layouts: it copies your own saved layout of every table (or only the tables you tick), so set your tables up the way you want first.

In the preset you choose who gets it: Everyone in the tenant, any groups, and any users. Weight settles a tie between two presets that reach someone at the same level - the lower weight wins, then the name. The preset lists its tables with what each shows (8 shown · 3 hidden · 1 pinned); remove a table there, or Re-capture from my layouts to copy in your current layouts again. Saving a preset, editing it and deleting it are recorded in the change log.

A preset is a default. Someone who has saved their own layout of a table keeps it. To bring people onto the preset anyway, use Reset personal layouts in the preset: it first says how many saved layouts of how many users across how many tables it would drop, and only then drops them - for the preset's tables and the people it reaches, nothing else. The reset is recorded in the change log on the preset.

Everyone keeps Reset in the Columns menu, which goes back to the default that applies to them; the menu names the preset that supplies it (Default: NOC view).

Only deployment administrators (the same people who can open Settings → Table layouts) can create, change, assign or reset presets; everyone else only receives them. A preset belongs to its tenant: it can only be given to that tenant's users, and it never applies in another tenant.

For scripts: GET /api/prefs/presets/ lists the tenant's presets and POST makes one (name, description, everyone, weight, user_ids, group_ids, and capture - "all" or a list of table ids); PATCH /api/prefs/presets/<id>/ changes one (the same fields, plus remove_tables), DELETE removes it, and POST /api/prefs/presets/<id>/reset/ with {"dry_run": true} counts what a reset would drop. A table's layout, GET /api/prefs/columns/<table>/, says where it came from (source: user, preset, default, tenant_forced or none) and names the preset behind the default (preset). The Preferences page reads all tables at once from GET /api/prefs/columns/, which gives the same source and preset per table, and shows a preset-supplied layout as Preset: .

How a layout is chosen

Danbyte shows the most specific layout that applies, in this order:

  1. a locked tenant default - it wins over everything;
  2. your own saved layout;
  3. a layout preset given to you;
  4. a layout preset given to one of your groups (lowest weight first, then name);
  5. a layout preset given to everyone;
  6. a tenant default published with Publish (above);
  7. the table's natural order.

When a table is locked, its column controls are disabled and a lock icon appears. A column added to Danbyte since a layout was saved takes its default, as in any saved layout: a new built-in column is shown, fields and custom fields stay hidden until a layout shows them. A preset's columns that a table no longer has are ignored.

What a saved layout records

A layout remembers the columns it knew about and which of them were hidden. A column you ticked stays shown; a column added to Danbyte after you saved takes its default - shown for new built-in columns (in their designed place), hidden for fields and custom fields.