X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/3cc81d1201ec9abfe818207c450df62232f5bce9..35895ee91c820680bb7df9696ab2e92525ead2ac:/sdk/python/arvados/util.py

diff --git a/sdk/python/arvados/util.py b/sdk/python/arvados/util.py
index 6c9822e9f0..a4b7e64a05 100644
--- a/sdk/python/arvados/util.py
+++ b/sdk/python/arvados/util.py
@@ -388,6 +388,68 @@ def list_all(fn, num_retries=0, **kwargs):
         offset = c['offset'] + len(c['items'])
     return items
 
+def keyset_list_all(fn, order_key="created_at", num_retries=0, ascending=True, **kwargs):
+    pagesize = 1000
+    kwargs["limit"] = pagesize
+    kwargs["count"] = 'none'
+    asc = "asc" if ascending else "desc"
+    kwargs["order"] = ["%s %s" % (order_key, asc), "uuid %s" % asc]
+    other_filters = kwargs.get("filters", [])
+
+    if "select" in kwargs and "uuid" not in kwargs["select"]:
+        kwargs["select"].append("uuid")
+
+    nextpage = []
+    tot = 0
+    expect_full_page = True
+    seen_prevpage = set()
+    seen_thispage = set()
+    lastitem = None
+    prev_page_all_same_order_key = False
+
+    while True:
+        kwargs["filters"] = nextpage+other_filters
+        items = fn(**kwargs).execute(num_retries=num_retries)
+
+        if len(items["items"]) == 0:
+            if prev_page_all_same_order_key:
+                nextpage = [[order_key, ">" if ascending else "<", lastitem[order_key]]]
+                prev_page_all_same_order_key = False
+                continue
+            else:
+                return
+
+        seen_prevpage = seen_thispage
+        seen_thispage = set()
+
+        for i in items["items"]:
+            # In cases where there's more than one record with the
+            # same order key, the result could include records we
+            # already saw in the last page.  Skip them.
+            if i["uuid"] in seen_prevpage:
+                continue
+            seen_thispage.add(i["uuid"])
+            yield i
+
+        firstitem = items["items"][0]
+        lastitem = items["items"][-1]
+
+        if firstitem[order_key] == lastitem[order_key]:
+            # Got a page where every item has the same order key.
+            # Switch to using uuid for paging.
+            nextpage = [[order_key, "=", lastitem[order_key]], ["uuid", ">" if ascending else "<", lastitem["uuid"]]]
+            prev_page_all_same_order_key = True
+        else:
+            # Start from the last order key seen, but skip the last
+            # known uuid to avoid retrieving the same row twice.  If
+            # there are multiple rows with the same order key it is
+            # still likely we'll end up retrieving duplicate rows.
+            # That's handled by tracking the "seen" rows for each page
+            # so they can be skipped if they show up on the next page.
+            nextpage = [[order_key, ">=" if ascending else "<=", lastitem[order_key]], ["uuid", "!=", lastitem["uuid"]]]
+            prev_page_all_same_order_key = False
+
+
 def ca_certs_path(fallback=httplib2.CA_CERTS):
     """Return the path of the best available CA certs source.
 
@@ -430,3 +492,28 @@ def get_config_once(svc):
     if not hasattr(svc, '_cached_config'):
         svc._cached_config = svc.configs().get().execute()
     return svc._cached_config
+
+def get_vocabulary_once(svc):
+    if not svc._rootDesc.get('resources').get('vocabularies', False):
+        # Old API server version, no vocabulary export endpoint
+        return {}
+    if not hasattr(svc, '_cached_vocabulary'):
+        svc._cached_vocabulary = svc.vocabularies().get().execute()
+    return svc._cached_vocabulary
+
+def trim_name(collectionname):
+    """
+    trim_name takes a record name (collection name, project name, etc)
+    and trims it to fit the 255 character name limit, with additional
+    space for the timestamp added by ensure_unique_name, by removing
+    excess characters from the middle and inserting an ellipse
+    """
+
+    max_name_len = 254 - 28
+
+    if len(collectionname) > max_name_len:
+        over = len(collectionname) - max_name_len
+        split = int(max_name_len/2)
+        collectionname = collectionname[0:split] + "…" + collectionname[split+over:]
+
+    return collectionname