From 7471ad1e74ef5b0d718398f453300df4a58d2eba Mon Sep 17 00:00:00 2001
From: Tim Holy <tim.holy@gmail.com>
Date: Mon, 20 Jul 2026 13:02:35 -0500
Subject: [PATCH] Fix Uniprot decompression and refresh EBI test accessions

Uniprot serves ID-mapping responses uncompressed unless the client negotiates
gzip, and does not always honor the request when it does. Requesting compression
and then unconditionally gunzipping the response therefore fails intermittently
with a ZlibError. Suppress the Accept-Encoding header via `decompress = false`
and read the body directly.

The EBI batch-query test used accession C3N734, which has since been withdrawn
from Uniprot, so the query returned three records instead of four. Reviewed
(SwissProt) accessions are not withdrawn the way unreviewed entries are; use
four GPCR entries already exercised elsewhere in the suite.

Assisted-by: Claude Opus 4.8 <noreply@anthropic.com>
---
 src/naming_conventions.jl | 11 +++++++----
 test/runtests.jl          |  4 +++-
 2 files changed, 10 insertions(+), 5 deletions(-)

diff --git a/src/naming_conventions.jl b/src/naming_conventions.jl
index e59dffe..4cfc77e 100644
--- a/src/naming_conventions.jl
+++ b/src/naming_conventions.jl
@@ -166,9 +166,12 @@ Check the status of a Uniprot ID mapping job. Returns `true` if the results are
 ready. Otherwise, returns the status object.
 """
 function map_uniprot_status(jobID)
-    resp = HTTP.get("https://rest.uniprot.org/idmapping/status/$jobID", ["Accept" => "application/json"]; decompress = true)
+    # `decompress = false` suppresses the `Accept-Encoding: gzip` request header, so the
+    # body arrives as plain text. Uniprot does not always honor a compression request, and
+    # unconditionally gunzipping the response fails whenever it returns one uncompressed.
+    resp = HTTP.get("https://rest.uniprot.org/idmapping/status/$jobID", ["Accept" => "application/json"]; decompress = false)
     if resp.status == 200
-        status = JSON3.read(String(HTTP.decode(resp)))
+        status = JSON3.read(String(resp.body))
         haskey(status, "results") && return true
         return status
     end
@@ -181,9 +184,9 @@ end
 Retrieve the results of a Uniprot ID mapping job.
 """
 function map_uniprot_retrieve(jobID)
-    resp = HTTP.get("https://rest.uniprot.org/idmapping/stream/$jobID", ["Accept" => "application/json"]; decompress = true)
+    resp = HTTP.get("https://rest.uniprot.org/idmapping/stream/$jobID", ["Accept" => "application/json"]; decompress = false)
     if resp.status == 200
-        return JSON3.read(String(HTTP.decode(resp)))
+        return JSON3.read(String(resp.body))
     end
     return nothing
 end
diff --git a/test/runtests.jl b/test/runtests.jl
index 8375643..4a120bc 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -418,7 +418,9 @@ using Test
                 end
             end
             @test startswith(query_ebi_proteins("Q7TQA6"; format=:fasta), ">sp|Q7TQA6")
-            @test count(==('>'), query_ebi_proteins(["C3N734", "H2C869", "Q3V4T1", "P20220"]; format=:fasta)) == 4
+            # Reviewed (SwissProt) accessions: unreviewed entries get withdrawn when
+            # the underlying genome annotation changes, silently shrinking the count.
+            @test count(==('>'), query_ebi_proteins(["P29274", "P29275", "P15409", "Q7TQA6"]; format=:fasta)) == 4
             q = query_ncbi(obj)
             @test q["total_count"] == 1
         end