Why does the Kernel shut down?

Question

I didn't see this coming:

ClearAll[randomTimeSeries]

randomTimeSeries[seed_ : RandomInteger[{10^6, 10^7}], m_ : 3, n_ : 30,iRandRng_ : {1, 7}] :=     
    Module[{ds, vs, junk, ts},
      {junk, {ts}} = Do[
        BlockRandom[
          ds = NestList[(DatePlus[#, -RandomInteger[iRandRng]] &), AbsoluteTime[], n - 1] // Reverse;
          vs = RandomReal[{0, 1}, n] // Accumulate;
          TimeSeries[vs, {ds}, ResamplingMethod -> {"Interpolation", 0}, MetaInformation -> {"seed" -> seed + j}] // Quiet /* Sow, 
            RandomSeeding -> seed + j], {j, 1, m}] // Reap;
      ts // (TimeSeriesThread[Mean, #] &)
 ]

Every single time I evaluate randomTimeSeries[RandomInteger[{10^6, 10^7}]] // Accumulate I get

1. Error messages about interpolation even though I am using Quiet
2. High memory and CPU usage and
3. The Kernel quits with messages about there being insufficient memory

I'm on Win10, using v.12.1.

Does anybody know whats going on ?

Answer 1

This is as designed. If we set

ts = randomTimeSeries[RandomInteger[{10^6, 10^7}]]

then RegularlySampledQ[ts] is False which means that calling Accumulate on it will first trigger TimeSeriesResample on ts with respect to the minimal time step:

In[3]:= ts["MinimumTimeIncrement"]

Out[3]= {0.00926208}

which is very small compared to the other time increments:

In[5]:= Differences[ts["Times"]] // MinMax

Out[5]= {0.00926208, 432000.}

and hence the huge memory usage and hang and/or crash.

If the objective is to accumulate values only then either call Accumulate on ts["Values"] or create new TimeSeries object that forces TemporalRegularity to True and then apply Accumulate:

In[6]:= ts1 = TimeSeries[ts, TemporalRegularity -> True];

In[9]:= (Accumulate[ts1] // Head) // AbsoluteTiming

Out[9]= {0.000467, TemporalData}