Add 'stat' command and per-key memory estimation to rdb-cli

[moticless]

Summary

Adds memory-oriented capabilities to rdb-cli, surfaced through its generated output:

1. A new stat command that prints a human-formatted, estimated memory-statistics report.
2. New per-key memory specifiers for print (%n encoding, %z estimated bytes, %g largest element), plus optional printf-style field widths.

Both share a single key-size estimator, so the per-key %z number and the stat aggregates are computed the same way.

stat — generated output

$ rdb-cli dump.rdb stat
rdb-cli dump.rdb stat
Statistics (Memory is estimated):
type             keys          items  items/key   volatile   expired       memory        avg    mem%
hash          1560278       81456780         52          0         0        6.8 G      4.6 K   74.8%
set           7651482       27952535          3          3         3        1.5 G      216 B   16.9%
list          6017862        7906342          1          0         0      719.9 M      125 B    7.7%
string         106018              -          -          2         2       63.5 M      628 B    0.7%
stream              1            948        948          0         0       22.3 K     22.3 K    0.0%
TOTAL        15335641      117316605                     5         5        9.1 G
volatile keys hold 2.0 K (0.0% of memory)

Top 10 keys by memory:
memory type        db        items   avg item      ttl  key
61.2 M set          0      1369515       46 B     none  App:AccessGroup:001
57.4 M set          0      1290404       46 B     none  App:EntityIndex:All
38.2 M hash         0       858814       46 B     none  App:EntityMetadata:002
...

Extending rdb-cli Usage:

FORMAT_OPTIONS ('stat'):
        -t, --top <N>      Show the top N keys by estimated memory
        -n, --now <SEC>    Reference Unix time for expiry evaluation (Default: now)

print — extended per-key memory output

$ rdb-cli dump.rdb print --key "db=%d,key=%k,type=%t,enc=%n,bytes=%z,largest=%g,count=%i"
db=0,key=mylist1,type=list,enc=quicklist,bytes=66,largest=2,count=1
db=0,key=lzf_compressed,type=string,enc=string,bytes=180,largest=118,count=0
...

New specifiers for --key:

Specifier	Meaning
%n	Encoding (e.g. quicklist, listpack, intset, skiplist)
%z	Estimated in-memory size in bytes
%g	Size in bytes of the largest element

Specifiers also accept an optional width, e.g. %-20k, %10z.

Implementation notes

• New handlersCommon module holds the shared key-size accumulator, the in-memory size estimator, and file/lifecycle plumbing; both the print and stat handlers compose RdbxHandlersBase and register their own data callbacks over it.
• The size estimate uses rough per-encoding overhead constants in the spirit of Redis MEMORY USAGE — calibrated so large keys land within ~25% of live MEMORY USAGE. It is explicitly an estimate.
• Shared JSON/string escaping moved to extCommon (rdbxOutputPlainEscaping).
• New public API: RDBX_createHandlersToStat(p, topN, nowSecs, outFilename).
• Tests added in test/test_rdb_to_mem_stat.c.

Status

Draft — opening for early feedback on the output format and on the memory-estimation heuristic.

🤖 Generated with Claude Code