Add getMeanLength() and report the weight-length defaults

[gustavdelius]

getMeanLength()

The counterpart of getMeanWeight(), sharing its man page under the new title
Calculate the mean size of the community. It divides the total length of all
individuals in the selected species and size range by their number:

$$\bar l = \frac{\sum_i \int N_i(w),(w/a_i)^{1/b_i},dw}{\sum_i \int N_i(w),dw}$$

The integral goes through sizeIntegral() with the species × size length matrix
as the weighting factor, so the size-range mask, the bin-averaging under
second_order_w(bin_average = TRUE) and the MizerParams/MizerSim shape
handling all come for free, and the bin integral is still performed in exactly
one place. Methods for both classes, plus the internal helper
length_at_size(), which errors if a/b are missing or NA as
get_size_range_array() does.

Reporting the weight-length defaults

Writing that showed a = 0.01 and b = 3 were being filled in silently.

At the point the default is filled in — species_params.data.frame() now
passes a message to set_species_param_default(), which is the existing
mechanism (signal_info(), severity = "info", level = 3), so the report
sits alongside the ones for h, gamma and ks:

i No `a` column so using a = 0.01 in w = a l^b, with w in g and l in cm.
i No `b` column so using the isometric default b = 3 in w = a l^b.

At the point the conversion happens — calc_selectivity() reports a gear
whose selectivity is set from a length while the species' a or b was
defaulted. That is the case where an invented weight-length relationship changes
the dynamics rather than just a diagnostic, so it is level = 1 (survives the
setting that silences the routine chatter) and unhandled = "show" (both
setFishing() and calc_selectivity() are exported and get called outside any
with_info_level()). It names only the affected species and only the parameter
that was actually missing. Whether a gear selects on length is read off the
selectivity function's formals — mizer hands a selectivity function the species
parameters only so it can convert between length and weight — so a custom
sel_func taking species_params is covered, and knife_edge /
sigmoid_weight never trigger it.

Why it is worth saying: a = 0.01, b = 3 with weights in grams and lengths in
cm is Fulton's K = 1, the textbook fusiform fish. The exponent is a good
default, but a follows body shape over about two orders of magnitude. Against
the twelve North Sea species shipped with mizer (fitted a = 0.001–0.010,
b = 2.94–3.43) the default gives lengths 1–42% short at a weight of 1 g — 35%
short for the eel-like Sandeel.

Documentation

• NEWS.md; new section in inst/skills/upgrade-mizer-code/references/mizer-3.4.md
  with two rows in the symptom index; getMeanLength() added to the
  analyse-and-plot skill and quick reference; inst/llms.txt / docs/llms.txt;
  guides regenerated with build_guides().
• Corrects inst/skills/build-model/SKILL.md, which said the defaults were
  0.006 and 3 — wrong since 0b5b661 introduced a = 0.01, and it feeds the
  published guide.

Testing

devtools::test() on this commit: FAIL 0 | WARN 0 | SKIP 49 | PASS 4937.
New tests in test-indicator_functions.R (MizerSim and MizerParams against
hand-rolled sums, the min_l/min_w equivalence, the a = b = 1 case where
mean length must equal mean weight, both error paths), test-second_order_summary.R
(the required bin_average = TRUE test), test-species_params.R and
test-setFishing.R (the two reports, and silence when the parameters are
supplied or the gear selects on weight).

🤖 Generated with Claude Code