Date: Mon, 15 Jul 2024 10:31:08 -0500
From: Nathan Bossart <nathandbossart@gmail.com>
To: Joseph Koshakow <koshy44@gmail.com>
Cc: Alexander Lakhin <exclusion@gmail.com>, Tom Lane <tgl@sss.pgh.pa.us>,
	PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>,
	Andres Freund <andres@anarazel.de>
Subject: Re: Remove dependence on integer wrapping
Message-ID: <ZpVAvLbtUPnK1JKe@nathan>
References: <ZmZbXK1CNMbkty0W@nathan>
 <618777.1717984674@sss.pgh.pa.us>
 <31ad2cd1-db94-bdb3-f91a-65ffdb4bef95@gmail.com>
 <CAAvxfHfqJca-WGvWJQ3ZAtt2po=mkQzU8r9Qjzom4YqxQi84oA@mail.gmail.com>
 <CAAvxfHdhEz4eF7z1wNn_4C4Gkt+e-h7OjzKUJXt7CE+CvubqcQ@mail.gmail.com>
 <CAAvxfHdze21+UZPWo7ePfH19-9hvXe_9SEZk4HABPHuzaUitRQ@mail.gmail.com>
 <CAAvxfHd22ujbY6uTraE+ZSTUN0xX0G-=NZcr0YnKSQuXthcM+g@mail.gmail.com>
 <CAAvxfHcDQEziyQ-ehv0Bg05OovGPGO95VoO7-DyxJx6Tkq85vA@mail.gmail.com>
 <ZpFeopUrhdB5GgNN@nathan>
 <CAAvxfHdX-t03LvxROxbyodBrtzBzZQEKdKB0FCpEr+p1JnFxsA@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: 
 <CAAvxfHdX-t03LvxROxbyodBrtzBzZQEKdKB0FCpEr+p1JnFxsA@mail.gmail.com>
Archived-At: <https://www.postgresql.org/message-id/ZpVAvLbtUPnK1JKe%40nathan>
Precedence: bulk

I took a closer look at 0002.

+    if (unlikely(isinf(f) || isnan(f)))
+        ereport(ERROR,
+                (errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
+                 errmsg("invalid float value")));
+
+    fresult = rint(f * c);

+    if (unlikely(f == 0.0))
+        ereport(ERROR,
+                (errcode(ERRCODE_DIVISION_BY_ZERO),
+                 errmsg("division by zero")));
+    if (unlikely(isinf(f) || isnan(f)))
+        ereport(ERROR,
+                (errcode(ERRCODE_NUMERIC_VALUE_OUT_OF_RANGE),
+                 errmsg("invalid float value")));
+
+    fresult = rint(c / f);

I'm curious why you aren't using float8_mul/float8_div here, i.e.,

	fresult = rint(float8_mul((float8) c, f));
	fresult = rint(float8_div((float8) c, f));

nitpick: I'd name the functions something like "cash_mul_float8" and
"cash_div_float8".  Perhaps we could also add functions like
"cash_mul_int64" and "cash_sub_int64" so that we don't need several copies
of the same "money out of range" ERROR.

-- 
nathan