Blank numeric fields parse to NaN not zero

1. What you'll see

   A total comes out wrong. Not visibly broken — just lower than it should be, or NaN in one cell of an otherwise correct table. It reproduces only for records where one numeric field was left blank, and every guard in the code looks right:

   let q = parseInt(row.Quantity);   // "" -> NaN
   let sum = q * row.Price;          // NaN
   return sum;                       // NaN reaches the page

   Code written against an older engine is the common source. Docly's numeric parsing has not always matched the spec, so a workspace can contain parse sites that were correct when they were written and are a source of NaN now.

2. What's actually happening

   parseInt("") and parseFloat("") return NaN, per the ECMAScript specification. So does parseInt(null). Once a NaN enters a calculation it propagates through every subsequent operation — NaN * 2, NaN + 10 and NaN / 3 are all NaN — so the failure surfaces a long way from the field that caused it.

   Number() is not a drop-in alternative, because it disagrees in both directions:

   Input	parseFloat	Number
   ""	NaN	0
   null	NaN	0
   "83 %"	83	NaN
   "12.5"	12.5	12.5

   That difference decides which one belongs where. A rendered table cell holds a formatted value like "83 %", which parseFloat reads as 83 and Number turns into NaN. A stored numeric field holds either a clean number or nothing, and Number("") is already 0, so on the server Number needs an isNaN guard only to catch genuine garbage.

3. What to do

   1. Guard at the parse site, not at the readout. This restores the intent the code was written with, rather than inventing a policy further down:

   let q = parseInt(row.Quantity);
   if (isNaN(q)) q = 0;

   2. A falsy check hides the problem rather than handling it. if (!sum) sum = 0; and sum || 0 are the usual reflexes, and both do catch NaN — which is exactly what makes them the wrong tool. The guard fires, the corrupted total is replaced with a plausible one, and the blank field that caused it leaves no trace. A wrong number hides far better than a visible NaN. Test for the condition you actually mean, with isNaN(), at the point where the value enters the calculation.

   3. Pick the function by side.

   Where	Use	Why
   Client-side, reading rendered cells	parseFloat	A cell holds "83 %"; Number gives NaN
   Server-side, reading stored fields	Number	Number("") and Number(null) are already 0

   4. When a total is wrong rather than missing, suspect a blank field before you suspect the arithmetic. A single un-guarded parse poisons every figure downstream of it, so the symptom appears in the summary row while the cause sits in one record.

   5. Check the behaviour rather than assuming it if you are working in a codebase old enough to predate the current engine. A throwaway page settles it in one request — see Use a scratch hash file to test Docly functions.

Last reviewed 21 Aug 2026