Re: [gpsd-users] Valid data in gps_data_t struct

[Gary E. Miller]

Yo Massimiliano!

On Sat, 25 May 2019 22:15:47 +0200
Massimiliano Fago <address@hidden> wrote:

> in some example I see this:
> 
> if (newdata->set & ONLINE_SET)
> (void)fprintf(stdout, "ONLINE: %lf\n", newdata->online);

Notice that is in the gpsd daemon, not in a client!

I assume you are working on a client, not on a driver for the daemon?

If you are working on a client then you need to look at libgps_json
and some of the example clients.

> if (newdata->set & DOP_SET)
> (void)fprintf(stdout,
> "DOP: satellites %d, pdop=%lf, hdop=%lf, vdop=%lf\n",
> newdata->satellites_used, newdata->dop.pdop,
> newdata->dop.hdop, newdata->dop.vdop);

Notice that there are seven different DOP in struct dop_t.  Not all
GPS provide all of them.  gpsd tries to fill in the missing ones, but may
not have sufficient data to do that.

> it is correct to check "newdata->set & DOP_SET" before use the data?

Too coarse a check.  Look at each one individually.

Look in gpsd_json.c starting at line 329:

    if (isfinite(datap->dop.xdop) != 0)
        str_appendf(reply, replylen, "\"xdop\":%.2f,", datap->dop.xdop);
    if (isfinite(datap->dop.ydop) != 0)
        str_appendf(reply, replylen, "\"ydop\":%.2f,", datap->dop.ydop);
[...]

Note the check for isfinite().  That is not a check for isnan().  A
subtle details that breaks a lot of software.

Here is an example, generic, but basically correct in any language
purporting IEEE 754 compliance:

    a = 1.0 / 0.0
    b = -1.0 / 0.0

You can divide 1.0, and -1.0, by 0.0 an infinite number of times.  So the
IEEE 754 Floating Point Math says the answers are:

    a == +INF
    b == -INF

+INF, and -INF, are NANs, but for weird historical reasons lost in
time they both test as:

    0 == isnan(+INF)
    0 == isnan(-INF)

Similar to:

    0 == isnan(1.0)
    0 == isnan(-1.0)

There are many different NANs and one must be careful using them.

In gpsd, what we want to know is if the value is a representable 
plain number.  isfinite() is the test for that.  Like this:

    0 == isfinite(+INF)
    0 == isfinite(-INF)
    1 == isfinite(1.0)
    1 == isfinite(-1.0)

Very few programmers understand IEEE 754.  Learning and using it as
intended can seriously speed up, and improve, floating point code.

> For example if my gps is not receving data, how I can know if pps
> subtracted themselves is valid?

The answer is in your question: if you are not receiving data then it is
not only not-valid, but non-existent.  The problem is that one person's
"fresh enough" is another persons "stale".  So check the fix time depending
on your freshness needs.

> The data is set to nan?

Mixing apples and oranges.  NAN is for floats and doubles.  timespec is
integers.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
        address@hidden  Tel:+1 541 382 8588

            Veritas liberabit vos. -- Quid est veritas?
    "If you can’t measure it, you can’t improve it." - Lord Kelvin

pgpBSegeXJHb7.pgp
Description: OpenPGP digital signature