From d4aed378cf9715efb69c9693528fa6ac0ee322a8 Mon Sep 17 00:00:00 2001
From: Alan Orth <alan.orth@gmail.com>
Date: Wed, 22 Mar 2023 12:16:56 +0300
Subject: [PATCH] Switch to pandas 2.0.0rc1

Seems to work fine with the new PyArrow datatypes.
---
 csv_metadata_quality/app.py | 3 +--
 pyproject.toml              | 2 +-
 2 files changed, 2 insertions(+), 3 deletions(-)

diff --git a/csv_metadata_quality/app.py b/csv_metadata_quality/app.py
index f5c7ceb..2e38308 100644
--- a/csv_metadata_quality/app.py
+++ b/csv_metadata_quality/app.py
@@ -73,8 +73,7 @@ def run(argv):
     # set the signal handler for SIGINT (^C)
     signal.signal(signal.SIGINT, signal_handler)
 
-    # Read all fields as strings so dates don't get converted from 1998 to 1998.0
-    df = pd.read_csv(args.input_file, dtype=str)
+    df = pd.read_csv(args.input_file, dtype_backend="pyarrow")
 
     # Check if the user requested to skip any fields
     if args.exclude_fields:
diff --git a/pyproject.toml b/pyproject.toml
index b31742a..a791ad3 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -12,7 +12,7 @@ csv-metadata-quality = 'csv_metadata_quality.__main__:main'
 
 [tool.poetry.dependencies]
 python = "^3.9"
-pandas = "^1.5.2"
+pandas = "2.0.0rc1"
 python-stdnum = "^1.18"
 requests = "^2.28.2"
 requests-cache = "^0.9.8"