Docs · Operations

Browsing

Walk the address space one level, all levels, or with continuation. browse() is the lowest-friction call; browseAll() and browseRecursive() handle paginated and tree traversals.

A Browse service call asks the server: "starting from this node, which references match my criteria and which target nodes do they point to?". The reply is a list of ReferenceDescription items — each describes one outgoing reference plus the (decoded) target node.

This library exposes three browse shapes — browse(), browseAll(), browseRecursive() — plus two pagination primitives. Pick one by intent:

You want	Use
One level, first page only	`browse()`
One level, all pages	`browseAll()`
Entire subtree	`browseRecursive()`
Manual continuation control	`browseWithContinuation()` + `browseNext()`

browse()

php examples/browse.php

foreach ($client->browse('i=85') as $ref) {
    printf(
        "%-30s %-12s %s\n",
        $ref->displayName->text,
        $ref->nodeClass->name,
        $ref->nodeId
    );
}

browse() returns the first page of references. For most servers, "a folder with a few children" fits in one page and there is nothing more to fetch. For wide nodes (a server with thousands of devices under one root), you will see only the first ~100–1000 entries; use browseAll() or browseWithContinuation().

Filters

Argument	Effect
`$direction`	`Forward` (default), `Inverse`, or `Both`. Inverse follows references pointing at the node.
`$includeSubtypes`	When `true` (default), references whose `ReferenceTypeId` is a subtype of the standard reference types are included. Set `false` to follow a single specific reference type.
`$nodeClassMask`	Bitmask filter on the target node's class. `0` = all. Combine `NodeClass` enum values with `\|`.

php filter to variables only

use PhpOpcua\Client\Types\NodeClass;

$variables = $client->browse(
    'ns=2;s=Devices',
    nodeClassMask: NodeClass::Variable->value
);

browseAll()

browseAll() follows continuation points until the server reports done. The signature mirrors browse():

php exhaustive single-level browse

$all = $client->browseAll('ns=2;s=Devices');   // every immediate child

Behind the scenes the client issues Browse, then BrowseNext on the continuation point repeatedly until the server returns no further items. Each follow-up call respects the same $useCache flag.

browseRecursive()

For tree walks, browseRecursive() traverses children-of-children up to a configurable depth, with built-in cycle detection:

php full subtree

$tree = $client->browseRecursive('ns=2;s=Devices', maxDepth: 3);

// $tree is a BrowseNode — each node has a reference and an array of children.
walk($tree, depth: 0);

function walk(BrowseNode $node, int $depth): void
{
    echo str_repeat('  ', $depth) . $node->reference->displayName->text . "\n";
    foreach ($node->getChildren() as $child) {
        walk($child, $depth + 1);
    }
}

$maxDepth = null falls back to the builder-wide default (setDefaultBrowseMaxDepth(), ships at 4). Cap it deliberately for large address spaces — an OPC UA server with 50 000 nodes will exhaust memory long before the network does.
Cycle detection tracks NodeIds already visited and short-circuits.
Internally, each level runs browseAll() and inherits its caching.

See Recipes · Browsing recursively for streaming alternatives.

Manual continuation

When you need finer control — incremental loading in a UI, time- bounded discovery — drop to the lower-level pair:

php manual pagination

$page = $client->browseWithContinuation('ns=2;s=Devices');

foreach ($page->references as $ref) {
    process($ref);
}

while ($page->continuationPoint !== null) {
    $page = $client->browseNext($page->continuationPoint);
    foreach ($page->references as $ref) {
        process($ref);
    }
}

browseWithContinuation() returns a BrowseResultSet with both the current page and an optional continuation point. browseNext() takes that token and returns the next page. The continuation point is opaque and unique to the originating call — pass it back unchanged.

Caching

By default every browse call caches its result keyed by endpoint URL, NodeId, direction, includeSubtypes, and nodeClassMask. Cache hits fire the CacheHit event, misses fire CacheMiss. Bypass per call with useCache: false; flush all entries with $client->flushCache() or invalidate one node with $client->invalidateCache($nodeId).

See Observability · Caching.

Browse events

Event	When
`NodeBrowsed`	After every successful browse — payload carries the node id, direction, and reference count
`CacheHit` / `CacheMiss`	Cache observation events