The Art of Visualization with Plotly

“”” ========================================================================= CRAFT — Plotly Visualization (Single Function) Context · Reference · Aesthetic · Framing · Telling (Storytelling) ========================================================================= FUNCTION SIGNATURE craft_plot(data, numeric, categorical, chart, title, label, color, shape) ARGUMENTS data pd.DataFrame numeric str | list[str] | None categorical str | list[str] | None chart str — see CHART MENU title str label dict — {col: ‘display name’} color str → column name : encodes that variable by color → palette name : applies that palette (e.g. ‘Set2’) → hex / named : single color (e.g. ‘#636EFA’, ‘red’) tuple → (col, palette): encode col + apply palette e.g. color=(‘sex’, ‘Set2’) shape str → column name : encodes that variable by marker symbol → symbol name : fixed symbol for all markers e.g. ‘circle’,’square’,’diamond’,’cross’,’x’, ‘triangle-up’,’triangle-down’,’star’,’pentagon’ tuple → (col, [sym1, sym2, …]): encode col + map symbols list length must match number of unique values in col e.g. shape=(‘smoker’, [‘diamond’, ‘circle’]) applies to: scatter, facet_scatter only CHART MENU Univariate numeric numeric=col histogram | box | violin | ecdf | kde Univariate categorical categorical=col bar | pie | donut Bivariate num × num numeric=[x, y] scatter | line | hexbin Bivariate num × cat numeric=col, categorical=col box | violin | strip | bar | stacked_bar | grouped_bar Bivariate cat × cat categorical=[c1, c2] heatmap | bar | stacked_bar | grouped_bar Multivariate (3+ variables, any mix) facet_histogram | facet_box | facet_scatter facet_bar | facet_violin | corr_heatmap ================================================================ “”” import math import numpy as np import pandas as pd import plotly.express as px import plotly.graph_objects as go from plotly.subplots import make_subplots # ── registries ──────────────────────────────────────────────────────────────── _QUALITATIVE = { p.lower(): p for p in dir(px.colors.qualitative) if not p.startswith(“_”) and isinstance(getattr(px.colors.qualitative, p), list) } _MARKER_SYMBOLS = { “circle”, “square”, “diamond”, “cross”, “x”, “triangle-up”, “triangle-down”, “triangle-left”, “triangle-right”, “pentagon”, “hexagon”, “star”, “hourglass”, “bowtie”, “asterisk”, “hash”, } # ── helpers ─────────────────────────────────────────────────────────────────── def _to_palette(val): “””Resolve a palette name string to a color list, or return None.””” if isinstance(val, str) and val.lower() in _QUALITATIVE: return getattr(px.colors.qualitative, _QUALITATIVE[val.lower()]) return None def _parse_color(color, df): “”” Parse color arg (str or tuple) into (color_col, disc_seq, cont_scale). color_col : column name to encode by color disc_seq : list of colors to apply cont_scale : continuous colorscale name “”” if color is None: return None, None, None # ── tuple: (col, palette) ───────────────────────────────────────────────── if isinstance(color, tuple): col, style = color pal = _to_palette(style) or (style if isinstance(style, list) else [style]) return col, pal, None # ── str: column name ────────────────────────────────────────────────────── if color in df.columns: return color, None, None # ── str: qualitative palette name ───────────────────────────────────────── pal = _to_palette(color) if pal: return None, pal, None # ── str: continuous colorscale ──────────────────────────────────────────── try: px.colors.get_colorscale(color) return None, None, color except Exception: pass # ── str: plain hex / named color ───────────────────────────────────────── return None, [color], None def _parse_shape(shape, df): “”” Parse shape arg (str or tuple) into (shape_col, symbol_map). shape_col : column name to encode by marker symbol symbol_map : fixed symbol str OR list of symbols per group “”” if shape is None: return None, None # ── tuple: (col, [sym1, sym2, …]) ────────────────────────────────────── if isinstance(shape, tuple): col, symbols = shape return col, symbols # symbols is a list matching group order # ── str: column name ────────────────────────────────────────────────────── if shape in df.columns: return shape, None # Plotly picks default symbols per group # ── str: fixed marker symbol ────────────────────────────────────────────── if shape.lower() in _MARKER_SYMBOLS: return None, shape # fixed symbol for all markers raise ValueError( f”shape='{shape}’ is not a column name or a known Plotly marker symbol.\n” f”Known symbols: {sorted(_MARKER_SYMBOLS)}” ) # ── main function ───────────────────────────────────────────────────────────── def craft_plot( data, numeric=None, categorical=None, chart=None, title=None, label=None, color=None, shape=None, ): df = data.copy() label = label or {} num_cols = [numeric] if isinstance(numeric, str) else list(numeric or []) cat_cols = [categorical] if isinstance(categorical, str) else list(categorical or []) color_col, disc_seq, cont_scale = _parse_color(color, df) shape_col, symbol_map = _parse_shape(shape, df) # convenience locals group_palette = disc_seq or px.colors.qualitative.Plotly single_clr = (disc_seq[0] if disc_seq else “#636EFA”) def _scatter_kw(x, y): “””px.scatter kwargs wiring color + shape.””” kw = dict(template=”plotly_white”, title=title, labels=label, x=x, y=y) if color_col: kw[“color”] = color_col if disc_seq: kw[“color_discrete_sequence”] = disc_seq if cont_scale: kw[“color_continuous_scale”] = cont_scale if shape_col: kw[“symbol”] = shape_col return kw def _apply_symbol(fig): “””Post-apply fixed or mapped symbols after px builds the figure.””” if symbol_map is None: return if isinstance(symbol_map, list): # map each trace (group) to its symbol in order for i, trace in enumerate(fig.data): trace.marker.symbol = symbol_map[i % len(symbol_map)] else: # fixed symbol for all traces fig.update_traces(marker_symbol=symbol_map) # ══════════════════════════════════════════════════════════════════════════ # UNIVARIATE NUMERIC # ══════════════════════════════════════════════════════════════════════════ if len(num_cols) == 1 and not cat_cols: col = num_cols[0] chart = chart or “histogram” if chart == “histogram”: fig = px.histogram(df, x=col, nbins=30, color_discrete_sequence=[single_clr], template=”plotly_white”, title=title, labels=label) elif chart == “box”: fig = px.box(df, y=col, points=”outliers”, color_discrete_sequence=[single_clr], template=”plotly_white”, title=title, labels=label) elif chart == “violin”: fig = px.violin(df, y=col, box=True, color_discrete_sequence=[single_clr], template=”plotly_white”, title=title, labels=label) elif chart == “ecdf”: fig = px.ecdf(df, x=col, markers=True, lines=True, color_discrete_sequence=[single_clr], template=”plotly_white”, title=title, labels=label) elif chart == “kde”: from scipy.stats import gaussian_kde vals = df[col].dropna().values xs = np.linspace(vals.min(), vals.max(), 300) fig = go.Figure(go.Scatter( x=xs, y=gaussian_kde(vals)(xs), mode=”lines”, line=dict(color=single_clr, width=2.5), )) fig.update_layout(template=”plotly_white”, title=title, xaxis_title=label.get(col, col), yaxis_title=”Density”) else: raise ValueError(“Univariate numeric → histogram | box | violin | ecdf | kde”) # ══════════════════════════════════════════════════════════════════════════ # UNIVARIATE CATEGORICAL # ══════════════════════════════════════════════════════════════════════════ elif not num_cols and len(cat_cols) == 1: col = cat_cols[0] counts = df[col].value_counts().reset_index() counts.columns = [col, “count”] chart = chart or “bar” if chart == “bar”: fig = px.bar(counts, x=col, y=”count”, color=col, color_discrete_sequence=group_palette, template=”plotly_white”, title=title, labels=label) elif chart in (“pie”, “donut”): fig = px.pie(counts, names=col, values=”count”, hole=(0.4 if chart == “donut” else 0), color_discrete_sequence=group_palette, template=”plotly_white”, title=title) else: raise ValueError(“Univariate categorical → bar | pie | donut”) # ══════════════════════════════════════════════════════════════════════════ # BIVARIATE NUM × NUM # ══════════════════════════════════════════════════════════════════════════ elif len(num_cols) == 2 and not cat_cols: x, y = num_cols chart = chart or “scatter” if chart == “scatter”: fig = px.scatter(df, **_scatter_kw(x, y)) _apply_symbol(fig) elif chart == “line”: fig = px.line(df.sort_values(x), x=x, y=y, template=”plotly_white”, title=title, labels=label) elif chart == “hexbin”: fig = px.density_heatmap(df, x=x, y=y, nbinsx=30, nbinsy=30, color_continuous_scale=cont_scale or “Viridis”, template=”plotly_white”, title=title, labels=label) else: raise ValueError(“Bivariate num×num → scatter | line | hexbin”) # ══════════════════════════════════════════════════════════════════════════ # BIVARIATE NUM × CAT # ══════════════════════════════════════════════════════════════════════════ elif len(num_cols) == 1 and len(cat_cols) == 1: num, cat = num_cols[0], cat_cols[0] chart = chart or “box” grp_col = color_col or cat if chart == “box”: fig = px.box(df, x=cat, y=num, color=grp_col, color_discrete_sequence=group_palette, template=”plotly_white”, title=title, labels=label) elif chart == “violin”: fig = px.violin(df, x=cat, y=num, color=grp_col, box=True, points=”outliers”, color_discrete_sequence=group_palette, template=”plotly_white”, title=title, labels=label) elif chart == “strip”: fig = px.strip(df, x=cat, y=num, color=grp_col, color_discrete_sequence=group_palette, template=”plotly_white”, title=title, labels=label) elif chart == “bar”: agg = df.groupby(cat)[num].agg([“mean”, “std”]).reset_index() agg.columns = [cat, “mean”, “std”] fig = px.bar(agg, x=cat, y=”mean”, error_y=”std”, color=cat, color_discrete_sequence=group_palette, template=”plotly_white”, title=title, labels={**label, “mean”: f”Mean of {num}”}) # ── NEW: stacked_bar (num × cat) ────────────────────────────────────── # Requires color_col to define the stack segments. # Aggregates `num` by (cat × color_col), stacks color_col within each # cat group. Falls back to a plain stacked bar if no color_col given. elif chart == “stacked_bar”: stack_by = color_col or cat # stack dimension agg = ( df.groupby([cat, stack_by])[num] .sum() .reset_index() ) fig = px.bar( agg, x=cat, y=num, color=stack_by, barmode=”stack”, color_discrete_sequence=group_palette, template=”plotly_white”, title=title, labels=label, ) # ── NEW: grouped_bar (num × cat) ───────────────────────────────────── # Requires color_col to define the sub-groups. # Aggregates `num` by (cat × color_col), places bars side-by-side. elif chart == “grouped_bar”: grp_by = color_col or cat # sub-group dimension agg = ( df.groupby([cat, grp_by])[num] .mean() .reset_index() ) fig = px.bar( agg, x=cat, y=num, color=grp_by, barmode=”group”, color_discrete_sequence=group_palette, template=”plotly_white”, title=title, labels=label, ) else: raise ValueError( “Bivariate num×cat → box | violin | strip | bar | stacked_bar | grouped_bar” ) # ══════════════════════════════════════════════════════════════════════════ # BIVARIATE CAT × CAT # ══════════════════════════════════════════════════════════════════════════ elif not num_cols and len(cat_cols) == 2: c1, c2 = cat_cols chart = chart or “heatmap” if chart == “heatmap”: ct = pd.crosstab(df[c1], df[c2]) fig = go.Figure(go.Heatmap( z=ct.values, x=list(ct.columns), y=list(ct.index), colorscale=cont_scale or “Blues”, text=ct.values, texttemplate=”%{text}”, )) fig.update_layout(template=”plotly_white”, title=title, xaxis_title=label.get(c2, c2), yaxis_title=label.get(c1, c1)) elif chart == “bar”: grp = df.groupby([c1, c2]).size().reset_index(name=”count”) fig = px.bar(grp, x=c1, y=”count”, color=c2, barmode=”group”, color_discrete_sequence=group_palette, template=”plotly_white”, title=title, labels=label) # ── NEW: stacked_bar (cat × cat) ───────────────────────────────────── # Cross-tabulates c1 × c2, stacks c2 categories within each c1 group. elif chart == “stacked_bar”: grp = df.groupby([c1, c2]).size().reset_index(name=”count”) fig = px.bar( grp, x=c1, y=”count”, color=c2, barmode=”stack”, color_discrete_sequence=group_palette, template=”plotly_white”, title=title, labels=label, ) # ── NEW: grouped_bar (cat × cat) ───────────────────────────────────── # Cross-tabulates c1 × c2, places c2 bars side-by-side within c1. elif chart == “grouped_bar”: grp = df.groupby([c1, c2]).size().reset_index(name=”count”) fig = px.bar( grp, x=c1, y=”count”, color=c2, barmode=”group”, color_discrete_sequence=group_palette, template=”plotly_white”, title=title, labels=label, ) else: raise ValueError( “Bivariate cat×cat → heatmap | bar | stacked_bar | grouped_bar” ) # ══════════════════════════════════════════════════════════════════════════ # MULTIVARIATE — T: Telling through facet grids # ══════════════════════════════════════════════════════════════════════════ elif len(num_cols) + len(cat_cols) >= 3: chart = chart or ( “facet_scatter” if len(num_cols) >= 3 and not cat_cols else “facet_violin” if num_cols and cat_cols else “facet_bar” ) def _grid(n): ncols = min(3, n) return math.ceil(n / ncols), ncols if chart == “facet_histogram”: nrows, ncols = _grid(len(num_cols)) fig = make_subplots(rows=nrows, cols=ncols, subplot_titles=[label.get(c, c) for c in num_cols]) for i, col in enumerate(num_cols): r, c = divmod(i, ncols) fig.add_trace(go.Histogram( x=df[col], nbinsx=25, name=label.get(col, col), marker_color=group_palette[i % len(group_palette)], showlegend=False, ), row=r + 1, col=c + 1) fig.update_layout(template=”plotly_white”, title=title) elif chart == “facet_box”: nrows, ncols = _grid(len(num_cols)) fig = make_subplots(rows=nrows, cols=ncols, subplot_titles=[label.get(c, c) for c in num_cols]) for i, col in enumerate(num_cols): r, c = divmod(i, ncols) fig.add_trace(go.Box( y=df[col], name=label.get(col, col), boxpoints=”outliers”, marker_color=group_palette[i % len(group_palette)], showlegend=False, ), row=r + 1, col=c + 1) fig.update_layout(template=”plotly_white”, title=title) elif chart == “facet_scatter”: kw = dict(template=”plotly_white”, title=title, labels=label) if color_col: kw[“color”] = color_col if disc_seq: kw[“color_discrete_sequence”] = disc_seq if cont_scale: kw[“color_continuous_scale”] = cont_scale if shape_col: kw[“symbol”] = shape_col fig = px.scatter_matrix(df, dimensions=num_cols, **kw) fig.update_traces(diagonal_visible=False, showupperhalf=False, marker=dict(size=4, opacity=0.65)) _apply_symbol(fig) elif chart == “facet_bar”: nrows, ncols = _grid(len(cat_cols)) fig = make_subplots(rows=nrows, cols=ncols, subplot_titles=[label.get(c, c) for c in cat_cols]) for i, col in enumerate(cat_cols): r, c = divmod(i, ncols) vc = df[col].value_counts() fig.add_trace(go.Bar( x=vc.index.tolist(), y=vc.values.tolist(), name=label.get(col, col), marker_color=group_palette[i % len(group_palette)], showlegend=False, ), row=r + 1, col=c + 1) fig.update_layout(template=”plotly_white”, title=title) elif chart == “facet_violin”: pairs = [(n, c) for n in num_cols for c in cat_cols] nrows, ncols = _grid(len(pairs)) titles = [f”{label.get(n,n)} by {label.get(c,c)}” for n, c in pairs] fig = make_subplots(rows=nrows, cols=ncols, subplot_titles=titles) for i, (num, cat) in enumerate(pairs): r, c = divmod(i, ncols) for j, grp in enumerate(df[cat].unique()): vals = df.loc[df[cat] == grp, num].dropna() fig.add_trace(go.Violin( y=vals, name=str(grp), box_visible=True, points=”outliers”, showlegend=(i == 0), line_color=group_palette[j % len(group_palette)], ), row=r + 1, col=c + 1) fig.update_layout(template=”plotly_white”, title=title) elif chart == “corr_heatmap”: corr = df[num_cols].corr() fig = go.Figure(go.Heatmap( z=corr.values, x=list(corr.columns), y=list(corr.index), colorscale=”RdBu”, zmid=0, zmin=-1, zmax=1, text=corr.round(2).values, texttemplate=”%{text}”, colorbar=dict(title=”r”), )) fig.update_layout(template=”plotly_white”, title=title) else: raise ValueError( “Multivariate → facet_histogram | facet_box | facet_scatter | ” “facet_bar | facet_violin | corr_heatmap” ) else: raise ValueError( “Cannot infer situation. Pass at least one column to numeric or categorical.” ) return fig

df = pd.read_csv(“\datapath.csv”) num_var1= “numeric_variable” num_var2= “numeric_variable” num_var3= “numeric_variable” cat_var1= “Categorical_variable” cat_var2= “Categorical_variable” cat_var3= “Categorical_variable” cat_var4= “Categorical_variable”

# Univariate numeric craft_plot(df, numeric=num_var1, chart=”histogram”, color=”#63AAAA”, title=” Histogram chart”).show() # Univariate categorical craft_plot(df, categorical=cat_var1, chart=”pie”, color=”Pastel”, title=” Pie chart”).show() # Bivariate num × num craft_plot(df, numeric=[num_var1, num_var2], chart=”scatter”, color=”black”, title=” Bivariate num × num chart”).show() # Bivariate num × cat craft_plot(df, numeric=num_var2, categorical=cat_var1, chart=”box”, color=”Set2″, title=”Bivariate num × cat chart”).show() # Bivariate cat × cat craft_plot(df, categorical=[cat_var1, cat_var2], chart=”stacked_bar”, title=”Bivariate cat × cat chart”).show() # Multivariate — grid of histograms (all numeric) craft_plot(df, numeric=[num_var1, num_var2, num_var3], chart=”facet_histogram”, color=”Plotly”, title=”Grid of histograms (all numeric)”).show() # Multivariate — scatter matrix (all numeric pairs) craft_plot(df, numeric=[num_var1, num_var2, num_var3], chart=”facet_scatter”, color=cat_var2, title=”Scatter matrix (all numeric pairs)”).show() # Multivariate — box grid (numeric × categorical) craft_plot(df, numeric=[num_var1, num_var2], categorical=[cat_var1, cat_var3], chart=”facet_violin”, color=”Set2″, title=”Box grid (numeric × categorical)”).show() # Multivariate — bar grid (all categorical) craft_plot(df, categorical=[cat_var1, cat_var2, cat_var3], chart=”facet_bar”, color=”Pastel”, title=” Bar grid (all categorical)”).show() # Multivariate — correlation heatmap craft_plot(df, numeric=[num_var1, num_var2, num_var3], chart=”corr_heatmap”, title=” Correlation Matrix”).show() # multivariate # color tuple: col + palette craft_plot(df, numeric=[num_var1, num_var2], chart=”scatter”, color=(cat_var1, “Set2″), shape=”diamond”, title=”Multivariate scatter plot”).show() # shape as column (Plotly picks symbols) craft_plot(df, numeric=[num_var1, num_var2], chart=”scatter”, color=(cat_var1, “Set2”), shape=(cat_var4,[“square”, “star”]), title=”Multivariate scatter plot”).show()