• Jonathan Hudson's avatar
    Add performance-analysis.md and sql-developers-skill-gap.md · 5266a805
    Jonathan Hudson authored
    
    
    Two repo-root analyses completing the decision-maker document set:
    
    performance-analysis.md - in-depth comparison vs hand-crafted stored
    procedures/triggers/views/functions across four axes (runtime,
    write-path, scale, operational), grounded in measured numbers from this
    machine: SQL generation for all 7 markets ~0.29s (the entire
    "abstraction cost", paid at compile time - generated SQL has ZERO
    runtime penalty), full 66-model/87-assertion offline pipeline 4.1s,
    unit tests 2.1s, emit-sql 2.4s. Covers the engine-fit split (where SQL
    Server wins by orders of magnitude and where columnar/MPP wins 10-100x),
    the synchronous trigger tax vs asynchronous CDC, assertions as one
    set-based scan vs per-row triggers, partition-pruned incrementals vs
    the growing nightly window, market-count scaling (Germany's marginal
    cost: two parallel queries), operational performance (idempotent
    reruns, deterministic resubmissions, the 4-second verification loop),
    an honest "where hand-crafted genuinely wins" section (freshness,
    point access, tiny scale, specialist single-query tuning) and a
    per-axis verdict table.
    
    sql-developers-skill-gap.md - the people side: a full skills inventory
    (transfers directly / needs adapting / genuinely new / becomes
    obsolete) with the headline that the gap is asymmetric - small in
    knowledge, meaningful in habit; the feared gap (JavaScript) is a week
    of string-building glue, while the real gaps are git-not-server,
    derived-not-maintained state, and data-not-procedure. Covers the
    psychological gap ("where is my SQL?" -> dataform-sql/ as the security
    blanket; loss-of-mastery framing), five failure patterns with their
    tripwires (hand-editing generated SQL, if(market) creep, harness
    bypass...), a staged learning path using this repo as the curriculum
    (data diffs in days, builder changes in 2-4 weeks, market ownership in
    ~2 months, the worked scenarios as katas), who thrives vs struggles,
    and organisational guidance (first tickets are data diffs; don't hire
    JS developers to bridge a regulatory-judgement gap).
    
    Both linked from the root README table; repo links validated.
    
    Co-Authored-By: default avatarClaude Fable 5 <noreply@anthropic.com>
    5266a805