From 5f9843f7c94befe4bacd39ea6e67dce92ebb75cd Mon Sep 17 00:00:00 2001
From: Max Gabrielsson <max@gabrielsson.com>
Date: Tue, 19 Nov 2024 13:59:04 +0100
Subject: [PATCH] fix tests, timestmap precision

---
 src/excel/xlsx/read_xlsx.cpp         | 25 ++++++++++++++++---------
 test/sql/excel/xlsx/limits.test      |  2 +-
 test/sql/excel/xlsx/read_sparse.test | 12 ++++++------
 3 files changed, 23 insertions(+), 16 deletions(-)

diff --git a/src/excel/xlsx/read_xlsx.cpp b/src/excel/xlsx/read_xlsx.cpp
index 041ef42..3a56669 100644
--- a/src/excel/xlsx/read_xlsx.cpp
+++ b/src/excel/xlsx/read_xlsx.cpp
@@ -372,23 +372,30 @@ static unique_ptr<GlobalTableFunctionState> InitGlobal(ClientContext &context, T
 
 int64_t ExcelToEpochUS(const double serial) {
 	// Convert to microseconds since epoch
-	static constexpr auto SECONDS_PER_DAY = 86400;
-	static constexpr auto MICROSECONDS_PER_SECOND = 1000000;
-	static constexpr auto DAYS_BETWEEN_1900_AND_1970 = 25569;
+	static constexpr auto SECONDS_PER_DAY = 86400UL;
+	static constexpr auto MICROSECONDS_PER_SECOND = 1000000UL;
+	static constexpr auto DAYS_BETWEEN_1900_AND_1970 = 25569UL;
 
 	// Excel serial is days since 1900-01-01
-	const auto days = serial - DAYS_BETWEEN_1900_AND_1970;
-	const auto seconds = days * SECONDS_PER_DAY;
-	const auto micros = seconds * MICROSECONDS_PER_SECOND;
+	const auto serial_days = serial;
+	auto serial_secs = serial_days * SECONDS_PER_DAY;
+
+	if (std::fabs(serial_secs - std::round(serial_secs)) < 1e-3) {
+		serial_secs = std::round(serial_secs);
+	}
+
+	const auto epoch_secs = serial_secs - (DAYS_BETWEEN_1900_AND_1970 * SECONDS_PER_DAY);
+	const auto epoch_micros = epoch_secs * MICROSECONDS_PER_SECOND;
 
 	// Clamp to the range. Theres not much we can do if the value is out of range
-	if (micros <= static_cast<double>(NumericLimits<int64_t>::Minimum())) {
+	if (epoch_micros <= static_cast<double>(NumericLimits<int64_t>::Minimum())) {
 		return NumericLimits<int64_t>::Minimum();
 	}
-	if (micros >= static_cast<double>(NumericLimits<int64_t>::Maximum())) {
+	if (epoch_micros >= static_cast<double>(NumericLimits<int64_t>::Maximum())) {
 		return NumericLimits<int64_t>::Maximum();
 	}
-	return static_cast<int64_t>(micros);
+
+	return static_cast<int64_t>(epoch_micros);
 }
 
 static void TryCast(XLSXGlobalState &state, bool ignore_errors, const idx_t col_idx, ClientContext &context,
diff --git a/test/sql/excel/xlsx/limits.test b/test/sql/excel/xlsx/limits.test
index ef67ee3..b11a6e4 100644
--- a/test/sql/excel/xlsx/limits.test
+++ b/test/sql/excel/xlsx/limits.test
@@ -16,4 +16,4 @@ TO '__TEST_DIR__/test_limit.xlsx' (FORMAT 'XLSX', sheet_name 'test');
 ----
 Invalid Input Error: XLSX: Sheet row limit of '1048576' rows exceeded!
  * XLSX files and compatible applications generally have a limit of '1048576' rows
- * You can export larger sheets on your own risk by setting the 'sheet_row_limit' parameter to a higher value
+ * You can export larger sheets at your own risk by setting the 'sheet_row_limit' parameter to a higher value
diff --git a/test/sql/excel/xlsx/read_sparse.test b/test/sql/excel/xlsx/read_sparse.test
index a966824..4390169 100644
--- a/test/sql/excel/xlsx/read_sparse.test
+++ b/test/sql/excel/xlsx/read_sparse.test
@@ -5,16 +5,16 @@ query I
 SELECT * FROM read_xlsx('test/data/xlsx/sparse.xlsx');
 ----
 
-query I
-DESCRIBE SELECT column_name, column_type FROM read_xlsx('test/data/xlsx/sparse.xlsx');
+query II
+SELECT column_name, column_type FROM (DESCRIBE FROM read_xlsx('test/data/xlsx/sparse.xlsx'));
 ----
-duckdb	DOUBLE
+duck	DOUBLE
 
 # But if we disable the header, we can read the cell
-query I
-DESCRIBE SELECT column_name, column_type FROM read_xlsx('test/data/xlsx/sparse.xlsx', header = false);
+query II
+SELECT column_name, column_type FROM (DESCRIBE FROM read_xlsx('test/data/xlsx/sparse.xlsx', header = false));
 ----
-R465
+R465	VARCHAR
 
 # Voila!
 query I