From 0d47d68f776fca66aec0ba32b5147a3d7484402b Mon Sep 17 00:00:00 2001
From: Nicola Tonellotto <tonellotto@users.noreply.github.com>
Date: Wed, 24 Apr 2024 13:03:30 +0200
Subject: [PATCH] Update tsv.py

The following code doesn't run correctly it the docstore is not already cache:

import ir_datasets
import more_itertools

dataset = ir_datasets.load('msmarco-passage')

for batch in more_itertools.chunked(dataset.docs_iter(), 8196*4):
    print(len(batch))

With this quick fix, it can.
---
 ir_datasets/formats/tsv.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/ir_datasets/formats/tsv.py b/ir_datasets/formats/tsv.py
index 4059e829..1d96dc90 100644
--- a/ir_datasets/formats/tsv.py
+++ b/ir_datasets/formats/tsv.py
@@ -26,6 +26,7 @@ def __next__(self):
                 self.stream = io.TextIOWrapper(self.ctxt.enter_context(self.dlc[self.stream_idx].stream()))
             else:
                 self.stream = io.TextIOWrapper(self.ctxt.enter_context(self.dlc.stream()))
+        line = ''
         while self.pos < self.start:
             line = self.stream.readline()
             if line != '\n':