review fixes

Signed-off-by: jorgee <jorge.ejarque@seqera.io>

Author: jorgee

plugins/nf-tower/src/main/io/seqera/tower/plugin/datalink/PagedDataLinkContent.groovy

@@ -89,7 +89,7 @@ class PagedDataLinkContent implements Iterable<DataLinkItem> {
                     current = items.iterator()
                     nextToken = page?.nextPageToken as String
                 } catch (IOException e) {
-                    throw new RuntimeException(e)
+                    throw new UncheckedIOException(e)
                 }
             }
             return true

plugins/nf-tower/src/main/io/seqera/tower/plugin/datalink/SeqeraDataLinkClient.groovy

@@ -33,8 +33,6 @@ import io.seqera.tower.model.DataLinkProvider
 import io.seqera.tower.plugin.TowerClient
 import nextflow.exception.AbortOperationException
 
-import java.nio.file.Path
-
 /**
  * Typed client for Seqera Platform data-link API endpoints.
  *
@@ -65,7 +63,8 @@ class SeqeraDataLinkClient {
     }
 
     /**
-     * Resolve a data-link providers in the given workspace.
+     * Distinct provider identifiers present in the workspace, sorted.
+     * The returned set is unmodifiable; memoized per workspace.
      */
     @Memoized
     Set<String> getDataLinkProviders(long workspaceId) {
@@ -75,12 +74,17 @@ class SeqeraDataLinkClient {
             final p = it.next().provider?.toString()
             if (p) providers.add(p)
         }
-        return providers
+        return Collections.unmodifiableSet(providers)
     }
 
     /**
      * Resolve a data-link by {@code (provider, name)} in the given workspace.
-     * Iterates the API's list endpoint lazily and short-circuits on first match.
+     * Iterates the API's list endpoint lazily (server-side filtered by {@code name})
+     * and short-circuits on first match.
+     *
+     * Memoized per {@code (workspaceId, provider, name)} tuple. Note: Groovy's
+     * {@code @Memoized} caches successful returns only — a path that repeatedly
+     * references a non-existent data-link re-runs the search each time.
      */
     @Memoized
     DataLinkDto getDataLink(long workspaceId, String provider, String name) {
@@ -177,7 +181,7 @@ class SeqeraDataLinkClient {
 
         private Iterator<DataLinkDto> current = Collections.<DataLinkDto>emptyIterator()
         private int offset = 0
-        private long total = -1L   // unknown until first fetch
+        private long total = -1L   // -1 = unknown; set only when the server reports totalSize
         private boolean exhausted = false
 
         DataLinkListIterator(TowerClient towerClient, String endpoint, long workspaceId, int pageSize, String search = null) {
@@ -212,8 +216,11 @@ class SeqeraDataLinkClient {
             final items = (json.dataLinks as List<Map>)?.collect { Map m -> mapDataLink(m) } ?: Collections.<DataLinkDto>emptyList()
             current = items.iterator()
             offset += items.size()
-            if (total < 0) total = (json.totalSize as Long) ?: 0L
-            if (items.isEmpty() || offset >= total) exhausted = true
+            // Record the server-reported total only if present (null/missing → leave as -1 and
+            // rely on an empty-page response to signal exhaustion)
+            if (total < 0 && json.totalSize != null) total = json.totalSize as Long
+            // Exhausted when: this page is empty, OR we've reached the known total
+            if (items.isEmpty() || (total >= 0 && offset >= total)) exhausted = true
         }
     }
 

plugins/nf-tower/src/main/io/seqera/tower/plugin/fs/SeqeraFileSystemProvider.groovy

@@ -197,8 +197,15 @@ class SeqeraFileSystemProvider extends FileSystemProvider {
 
         final source = entries
         return new DirectoryStream<Path>() {
+            private boolean iteratorCalled = false
             @Override
             Iterator<Path> iterator() {
+                // NIO contract: DirectoryStream.iterator() may be called at most once.
+                // For data-link listings a second iteration would also re-fetch pages 2+
+                // (needlessly doubling API calls), so enforcing the contract is a win.
+                if (iteratorCalled)
+                    throw new IllegalStateException("DirectoryStream.iterator() may be called at most once")
+                iteratorCalled = true
                 final inner = source.iterator()
                 if (!filter) return inner
                 return new FilteredIterator<Path>(inner, filter)

plugins/nf-tower/src/main/io/seqera/tower/plugin/fs/handler/DataLinksResourceHandler.groovy

@@ -60,7 +60,10 @@ class DataLinksResourceHandler implements ResourceTypeHandler {
     private final HttpClient httpClient
 
     DataLinksResourceHandler(SeqeraFileSystem fs, SeqeraDataLinkClient client) {
-        this(fs, client, HttpClient.newBuilder().connectTimeout(Duration.ofSeconds(10)).build())
+        this(fs, client, HttpClient.newBuilder()
+                .connectTimeout(Duration.ofSeconds(10))
+                .followRedirects(HttpClient.Redirect.NORMAL)
+                .build())
     }
 
     /** Test-only constructor to inject a mock {@link HttpClient}. */
@@ -111,8 +114,15 @@ class DataLinksResourceHandler implements ResourceTypeHandler {
         if (p.cachedAttributes) return p.cachedAttributes
         final workspaceId = fs.resolveWorkspaceId(p.org, p.workspace)
         final trail = p.trail
-        if (trail.size() < 2) {
-            // data-links/ or data-links/<provider> — always directory
+        if (trail.isEmpty()) {
+            // data-links/ — always a directory
+            return new SeqeraFileAttributes(true)
+        }
+        if (trail.size() == 1) {
+            // data-links/<provider> — validate the provider has at least one data-link
+            final providers = client.getDataLinkProviders(workspaceId)
+            if (!providers.contains(trail[0]))
+                throw new NoSuchFileException(p.toString(), null, "No data-links for provider '${trail[0]}' in workspace '${p.workspace}'")
             return new SeqeraFileAttributes(true)
         }
         final dl = client.getDataLink(workspaceId, trail[0], trail[1])

plugins/nf-tower/src/test/io/seqera/tower/plugin/datalink/SeqeraDataLinkClientTest.groovy

@@ -113,6 +113,135 @@ class SeqeraDataLinkClientTest extends Specification {
         !client.listDataLinks(10L).hasNext()
     }
 
+    // ---- getDataLink ----
+
+    def "getDataLink uses server-side search filter and returns first matching provider"() {
+        given:
+        def body = JsonOutput.toJson([dataLinks: [
+                [id: 'dl-1', name: 'inputs', provider: 'google'],
+                [id: 'dl-2', name: 'inputs', provider: 'aws']
+        ], totalSize: 2])
+        def tc = tower()
+        tc.sendApiRequest("${EP}/data-links?workspaceId=10&max=100&offset=0&search=inputs") >> ok(body)
+        def client = new SeqeraDataLinkClient(tc)
+
+        when:
+        def dl = client.getDataLink(10L, 'aws', 'inputs')
+
+        then:
+        dl.id == 'dl-2'
+        dl.provider.toString() == 'aws'
+    }
+
+    def "getDataLink throws NoSuchFileException when no matching (provider, name) is found"() {
+        given:
+        def body = JsonOutput.toJson([dataLinks: [
+                [id: 'dl-1', name: 'inputs', provider: 'google']
+        ], totalSize: 1])
+        def tc = tower()
+        tc.sendApiRequest("${EP}/data-links?workspaceId=10&max=100&offset=0&search=inputs") >> ok(body)
+        def client = new SeqeraDataLinkClient(tc)
+
+        when:
+        client.getDataLink(10L, 'aws', 'inputs')
+
+        then:
+        thrown(NoSuchFileException)
+    }
+
+    def "getDataLink memoizes successful lookups"() {
+        given:
+        def body = JsonOutput.toJson([dataLinks: [
+                [id: 'dl-1', name: 'inputs', provider: 'aws']
+        ], totalSize: 1])
+        def tc = tower()
+        def client = new SeqeraDataLinkClient(tc)
+
+        when:
+        def a = client.getDataLink(10L, 'aws', 'inputs')
+        def b = client.getDataLink(10L, 'aws', 'inputs')
+
+        then:
+        1 * tc.sendApiRequest("${EP}/data-links?workspaceId=10&max=100&offset=0&search=inputs") >> ok(body)
+        a.is(b)
+    }
+
+    // ---- getDataLinkProviders ----
+
+    def "getDataLinkProviders returns distinct sorted providers across all pages"() {
+        given:
+        def p1 = JsonOutput.toJson([dataLinks: [
+                [id: 'dl-1', name: 'a', provider: 'aws'],
+                [id: 'dl-2', name: 'b', provider: 'google']
+        ], totalSize: 3])
+        def p2 = JsonOutput.toJson([dataLinks: [
+                [id: 'dl-3', name: 'c', provider: 'aws']
+        ], totalSize: 3])
+        def tc = tower()
+        tc.sendApiRequest("${EP}/data-links?workspaceId=10&max=100&offset=0") >> ok(p1)
+        tc.sendApiRequest("${EP}/data-links?workspaceId=10&max=100&offset=2") >> ok(p2)
+        def client = new SeqeraDataLinkClient(tc)
+
+        when:
+        def providers = client.getDataLinkProviders(10L)
+
+        then:
+        providers as List == ['aws', 'google']
+    }
+
+    def "getDataLinkProviders memoizes the result"() {
+        given:
+        def body = JsonOutput.toJson([dataLinks: [
+                [id: 'dl-1', name: 'a', provider: 'aws']
+        ], totalSize: 1])
+        def tc = tower()
+        def client = new SeqeraDataLinkClient(tc)
+
+        when:
+        def a = client.getDataLinkProviders(10L)
+        def b = client.getDataLinkProviders(10L)
+
+        then:
+        1 * tc.sendApiRequest("${EP}/data-links?workspaceId=10&max=100&offset=0") >> ok(body)
+        a.is(b)
+    }
+
+    def "getDataLinkProviders returns an unmodifiable Set"() {
+        given:
+        def body = JsonOutput.toJson([dataLinks: [
+                [id: 'dl-1', name: 'a', provider: 'aws']
+        ], totalSize: 1])
+        def tc = tower()
+        tc.sendApiRequest(_) >> ok(body)
+        def client = new SeqeraDataLinkClient(tc)
+
+        when:
+        client.getDataLinkProviders(10L).add('gcs')
+
+        then:
+        thrown(UnsupportedOperationException)
+    }
+
+    // ---- listDataLinks pagination robustness ----
+
+    def "listDataLinks keeps paginating when totalSize is absent until an empty page"() {
+        given:
+        def p1 = JsonOutput.toJson([dataLinks: [[id: 'dl-1', name: 'a', provider: 'aws']]]) // no totalSize
+        def p2 = JsonOutput.toJson([dataLinks: [[id: 'dl-2', name: 'b', provider: 'aws']]]) // no totalSize
+        def p3 = JsonOutput.toJson([dataLinks: []]) // empty page → exhausted
+        def tc = tower()
+        tc.sendApiRequest("${EP}/data-links?workspaceId=10&max=100&offset=0") >> ok(p1)
+        tc.sendApiRequest("${EP}/data-links?workspaceId=10&max=100&offset=1") >> ok(p2)
+        tc.sendApiRequest("${EP}/data-links?workspaceId=10&max=100&offset=2") >> ok(p3)
+        def client = new SeqeraDataLinkClient(tc)
+
+        when:
+        def list = drain(client.listDataLinks(10L))
+
+        then:
+        list*.id == ['dl-1', 'dl-2']
+    }
+
     // ---- getContent ----
 
     def "getContent on a sub-path uses /browse/{path}"() {

plugins/nf-tower/src/test/io/seqera/tower/plugin/fs/SeqeraFileSystemProviderTest.groovy

@@ -452,4 +452,36 @@ class SeqeraFileSystemProviderTest extends Specification {
         def ex = thrown(NoSuchFileException)
         ex.reason?.contains('Unsupported resource type')
     }
+
+    def "readAttributes short-circuits when the SeqeraPath carries cachedAttributes (no API call)"() {
+        given: 'a provider with a fresh filesystem and a path carrying pre-resolved attrs'
+        def tc = spyTower()
+        def fs = buildFs(tc)
+        def attrs = new SeqeraFileAttributes(999L, java.time.Instant.EPOCH, java.time.Instant.EPOCH, 'cached-key')
+        def path = new SeqeraPath(fs, 'seqera://acme/research/datasets/samples').resolveWithAttributes('nested', attrs)
+
+        when:
+        def got = fs.provider().readAttributes(path, java.nio.file.attribute.BasicFileAttributes)
+
+        then: 'no workspace-cache load and no dataset/browse API calls were issued'
+        0 * tc.sendApiRequest(_)
+        got === attrs
+    }
+
+    def "newDirectoryStream.iterator() throws IllegalStateException on a second call"() {
+        given:
+        def tc = spyTower()
+        tc.sendApiRequest("${ENDPOINT}/user-info") >> ok(userInfoJson())
+        tc.sendApiRequest("${ENDPOINT}/user/42/workspaces") >> ok(workspacesJson())
+        def fs = buildFs(tc)
+        def wsPath = new SeqeraPath(fs, 'seqera://acme/research')
+        def stream = fs.provider().newDirectoryStream(wsPath, null)
+
+        when:
+        stream.iterator()
+        stream.iterator()
+
+        then:
+        thrown(IllegalStateException)
+    }
 }

plugins/nf-tower/src/test/io/seqera/tower/plugin/fs/handler/DataLinksResourceHandlerTest.groovy

@@ -271,6 +271,33 @@ class DataLinksResourceHandlerTest extends Specification {
         !attr.regularFile
     }
 
+    def "readAttributes at data-links/<provider>/ reports directory when provider exists"() {
+        given:
+        def path = new SeqeraPath(fs, 'seqera://acme/research/data-links/aws')
+
+        when:
+        def attr = handler.readAttributes(path)
+
+        then:
+        1 * fs.resolveWorkspaceId(_, _) >> 10L
+        1 * client.getDataLinkProviders(10L) >> (['aws', 'google'] as Set)
+        attr.directory
+    }
+
+    def "readAttributes at data-links/<provider>/ throws when the provider has no data-links"() {
+        given:
+        def path = new SeqeraPath(fs, 'seqera://acme/research/data-links/azure')
+
+        when:
+        handler.readAttributes(path)
+
+        then:
+        1 * fs.resolveWorkspaceId(_, _) >> 10L
+        1 * client.getDataLinkProviders(10L) >> (['aws'] as Set)
+        def ex = thrown(NoSuchFileException)
+        ex.reason?.contains("No data-links for provider 'azure'")
+    }
+
     def "readAttributes at data-link root reports directory"() {
         given:
         def path = new SeqeraPath(fs, 'seqera://acme/research/data-links/aws/inputs')

specs/260422-seqera-datalinks-fs/spec.md

@@ -20,6 +20,7 @@
 - Q: Which DTOs are introduced by this feature? → A: None. All types are reused from the `io.seqera:tower-api:1.121.0` dependency (`DataLinkDto`, `DataLinkItem`, `DataLinkProvider`, `DataLinkCredentials`, `DataLinkContentResponse`, `DataLinkDownloadUrlResponse`, etc.). A plugin-local `PagedDataLinkContent` holder class wraps the eager-first-page + lazy-pagination behavior but holds only tower-api types.
 - Q: Is browse-per-file supported by the Platform API? → A: Yes. `GET /data-links/{id}/browse/{path}` works for both directories and files, so `readAttributes` on any path is a single targeted call — no parent-browse-and-filter, no N+1 problem.
 - Q: How are paginated Platform responses returned to callers? → A: Streaming. The workspace data-link list (`GET /data-links`) returns an `Iterator<DataLinkDto>` that fetches offsets on demand. The browse endpoint returns a `PagedDataLinkContent` that loads the first page eagerly (so `readAttributes` can inspect it without iterating) and fetches subsequent pages lazily as the iterator advances. The handler layer exposes `Iterable<Path>` to the NIO `DirectoryStream`; no full materialization of listings in memory.
+- Q: What convenience methods does the client expose on top of the raw list endpoint? → A: Two memoized helpers — `getDataLink(ws, provider, name)` uses the server-side `&search=<name>` filter and returns the first match (throws `NoSuchFileException` on miss); `getDataLinkProviders(ws)` returns the sorted set of distinct providers present in the workspace. Both are memoized per-arguments within a single `SeqeraDataLinkClient` instance.
 - Q: How are attributes discovered after a listing? → A: When `newDirectoryStream` yields a child path, the handler attaches the per-item attributes (size for files, directory marker for folders) to the `SeqeraPath` via an optional cache field. A subsequent `readAttributes` on that path returns the cached value without any additional Platform API call. Paths parsed from URIs (no prior listing) fall back to the live browse endpoint.
 - Q: How are cloud credentials for the underlying bucket/prefix selected? → A: The Platform's `DataLinkDto.credentials` list associates one or more credential records with a data-link. The plugin forwards the first credential's ID as the `credentialsId` query parameter on browse and download-URL requests, when present. If the data-link has no associated credentials, the parameter is omitted and the Platform uses its default resolution.
 - Q: Which provider-segment value appears in user-visible paths? → A: The lowercase value of the `DataLinkProvider` enum, as exposed by its `toString()` (e.g. `aws`, `google`, `azure`). This matches the Platform UI.