Tableplot Gaps

Issues and missing features encountered while translating fpp3 examples.

Discovered via fpp3 Chapter 2

Numeric columns treated as continuous color scales

• Context: §2.4/2.5 seasonal plots — :=color "MonthNum" where MonthNum is long
• Problem: Tableplot treats numeric columns as continuous gradients, not discrete categories
• Workaround: Convert to string before plotting (mapv str)
• Desired: Option to force categorical treatment, e.g. :=color-type :nominal

No seasonal plot helper

• Context: §2.4 gg_season() equivalent
• Problem: Building a seasonal overlay plot requires manual field extraction + color grouping
• Note: Partially addressed by add-time-columns in tablecloth.time. A layer-seasonal in tableplot would complete it.

No subseries plot / faceting by field

• Context: §2.5 gg_subseries() equivalent
• Problem: No built-in faceting support (one panel per month, with mean line per panel)
• Workaround: Built Plotly subplots manually:
  • Manual axis domain calculations (xaxis, xaxis2, ... with domain arrays)
  • Manual trace assignment ({:xaxis "x3" :yaxis "y3"})
  • Manual shared y-axis range calculation and application
  • Manual mean line traces (one horizontal line per facet)
  • ~80 lines of layout/trace construction code
• Desired: Something like:

  (-> ds
      (plotly/layer-line {:=x "Year" :=y "Cost"})
      (plotly/facet-wrap {:=facet "MonthNum" :=ncol 12})
      (plotly/add-hline {:=y :mean :=per-facet true}))  ; or similar
  

• Note: This was the most labor-intensive plot in Chapter 2. The time-series concepts (subseries decomposition, monthly means) were trivial — the work was entirely Plotly layout gymnastics. Faceting would be high-value for tableplot.

No reference line helpers (hline, vline, abline)

• Context: §2.5 mean lines per facet, §2.8 ACF significance bounds
• Problem: Adding horizontal/vertical reference lines requires manual trace construction
• Workaround: Created additional {:type "scatter" :mode "lines" :y [mean mean]} traces
• Desired: plotly/add-hline, plotly/add-vline helpers that work with facets

No scatterplot matrix

• Context: §2.6 GGally::ggpairs() equivalent
• Problem: No built-in pairwise scatterplot grid
• Workaround: Manual composition of individual scatterplots

No lag plot helper

• Context: §2.7 gg_lag() equivalent
• Problem: Plotting y_t vs y_{t-k} requires manual column construction
• Note: Could be a tablecloth.time function (lag-column) rather than a tableplot feature

No ACF bar plot with significance bounds

• Context: §2.8 ACF correlogram
• Problem: No built-in way to overlay horizontal reference lines (±1.96/√T bounds) on a bar chart
• Workaround: Added bound columns to dataset, but can't render them as horizontal lines

No per-trace customization or post-processing hook

• Context: §2.4 seasonal plots with 700+ traces (one per day)
• Problem: Need to hide legend, set custom colors by year, thin line widths — but tableplot only exposes uniform options (:=mark-color applies to ALL traces)
• Workaround: Call plotly/plot to force evaluation, then post-process the :data traces manually
• Desired: One or more of:

  ;; Option A: per-trace transform function
  {:=trace-fn (fn [trace] (assoc trace :showlegend false))}
  
  ;; Option B: full spec post-process hook
  {:=post-process (fn [spec] (update spec :data ...))}
  
  ;; Option C: more granular built-in options
  {:=showlegend false        ; per-trace
   :=line-width 0.3          ; already have :=mark-size, but not for lines
   :=color-fn #(case ...)}   ; function instead of just column + palette
  

• Note: plotly/plot is the official API for forcing evaluation, which is good. But the pattern of "tableplot gets 90% there, need to tweak output" seems common enough to warrant a hook.

---

Template for future entries

[Short title]

• Context: fpp3 §X.Y, [R function]
• Problem: [What's missing or broken]
• Workaround: [What we did instead]
• Desired: [What the ideal API would look like]