Skip to content

normalizer_utils

knime2py.nodes.normalizer_utils

Shared helpers for Normalizer-based nodes.

Contains
  • NormalizerSettings dataclass
  • settings.xml parsers
  • code-emission utilities that normalize numeric columns

parse_normalizer_settings(node_dir)

Parse Normalizer settings with disambiguation
  • Normalization mode: model/entry[@key='mode'] (or 'normalizationMethod'), but NOT inside dataColumnFilterConfig / data-column-filter.
  • new-min/new-max (or newMin/newMax) also taken outside the filter-config subtree.
  • Column filter: ONLY read manualFilter/manuallyDeselected (fallback: excluded_names).
Source code in src/knime2py/nodes/normalizer_utils.py 
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
def parse_normalizer_settings(node_dir: Optional[Path]) -> NormalizerSettings:
    """
    Parse Normalizer settings with disambiguation:
      - Normalization mode: model/entry[@key='mode'] (or 'normalizationMethod'), but NOT inside
        dataColumnFilterConfig / data-column-filter.
      - new-min/new-max (or newMin/newMax) also taken outside the filter-config subtree.
      - Column filter: ONLY read manualFilter/manuallyDeselected (fallback: excluded_names).
    """
    if not node_dir:
        return NormalizerSettings()
    sp = node_dir / "settings.xml"
    if not sp.exists():
        return NormalizerSettings()

    root = ET.parse(str(sp), parser=XML_PARSER).getroot()
    model = first_el(root, ".//*[local-name()='config' and @key='model']")
    if model is None:
        return NormalizerSettings()

    norm_mode = first(
        model,
        ".//*[local-name()='entry' and @key='normalizationMethod'"
        " and not(ancestor::*[local-name()='config' and (@key='dataColumnFilterConfig' or @key='data-column-filter')])"
        "]/@value",
    )

    if not norm_mode:
        norm_mode = first(
            model,
            ".//*[local-name()='entry' and @key='mode'"
            " and not(ancestor::*[local-name()='config' and (@key='dataColumnFilterConfig' or @key='data-column-filter')])"
            "]/@value",
        )

    raw_mode = (norm_mode or "MINMAX").strip()
    normalized = raw_mode.upper().replace("-", "").replace("_", "")
    mode = normalized
    allowed_modes = {"MINMAX", "ZSCORE", "DECIMALSCALING"}
    if mode not in allowed_modes:
        numeric_map = {
            "0": "ZSCORE",
            "1": "MINMAX",
            "2": "ZSCORE",
            "3": "DECIMALSCALING",
        }
        mapped = numeric_map.get(raw_mode.strip())
        if mapped:
            mode = mapped
        else:
            mode = "MINMAX"

    def _float_entry_excl(keys: List[str], default: float) -> float:
        """Retrieve a float entry from the model, excluding certain subtrees."""
        for k in keys:
            v = first(
                model,
                f".//*[local-name()='entry' and @key='{k}']"
                " [not(ancestor::*[local-name()='config' and (@key='dataColumnFilterConfig' or @key='data-column-filter')])]/@value",
            )
            if v is not None:
                try:
                    return float(v)
                except Exception:
                    pass
        return default

    new_min = _float_entry_excl(["new-min", "newMin"], 0.0)
    new_max = _float_entry_excl(["new-max", "newMax"], 1.0)

    excludes: List[str] = []
    dcf = first_el(
        model,
        ".//*[local-name()='config' and (@key='dataColumnFilterConfig' or @key='data-column-filter')]",
    )
    if dcf is not None:
        man_desel = first_el(
            dcf,
            ".//*[local-name()='config' and @key='manualFilter']"
            "/*[local-name()='config' and @key='manuallyDeselected']",
        )
        if man_desel is not None:
            excludes.extend(_collect_numeric_name_entries(man_desel))
        else:
            exc_old = first_el(dcf, ".//*[local-name()='config' and @key='excluded_names']")
            if exc_old is not None:
                excludes.extend(_collect_numeric_name_entries(exc_old))

    excludes = list(dict.fromkeys([c for c in excludes if c]))

    # Optional explicit columns list
    columns: List[str] = []
    cols_cfg = first_el(
        model,
        ".//*[local-name()='config' and @key='columns']",
    )
    if cols_cfg is not None:
        columns = _collect_numeric_name_entries(cols_cfg)

    # Flag for "all numeric columns"
    all_numeric_entry = first(
        model,
        ".//*[local-name()='entry' and @key='all_numeric_columns_used']/@value",
    )
    all_numeric = str(all_numeric_entry or "").strip().lower() == "true"

    return NormalizerSettings(
        mode=mode,
        new_min=new_min,
        new_max=new_max,
        excludes=excludes,
        columns=columns,
        all_numeric=all_numeric,
    )

emit_normalize_code(cfg, bundle_var='bundle')

Emit Python code lines that normalize numeric columns and capture stats.

Returns:

Type Description
List[str]

List[str]: Lines of code to append to the generated body.
Source code in src/knime2py/nodes/normalizer_utils.py 
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
def emit_normalize_code(cfg: NormalizerSettings, bundle_var: str = "bundle") -> List[str]:
    """
    Emit Python code lines that normalize numeric columns and capture stats.

    Returns:
        List[str]: Lines of code to append to the generated body.
    """
    mode = (cfg.mode or "MINMAX").upper()
    lines: List[str] = []
    lines.append("out_df = df.copy()")
    lines.append(f"{bundle_var} = {{")
    lines.append(f"    'mode': {repr(mode)},")
    lines.append(f"    'new_min': {cfg.new_min},")
    lines.append(f"    'new_max': {cfg.new_max},")
    lines.append(f"    'excludes': {repr(cfg.excludes)},")
    lines.append("    'columns': [],")
    lines.append("    'stats': {},")
    lines.append("}")
    lines.append("all_cols = out_df.columns.tolist()")
    if cfg.columns:
        lines.append(f"cand_cols = [c for c in {repr(cfg.columns)} if c in all_cols]")
    elif cfg.all_numeric:
        lines.append("cand_cols = out_df.select_dtypes(include=['number']).columns.tolist()")
    else:
        lines.append("cand_cols = all_cols")
    if cfg.excludes:
        exc_list = ", ".join(repr(c) for c in cfg.excludes)
        lines.append(f"exclude_cols = [{exc_list}]")
        lines.append("cand_cols = [c for c in cand_cols if c not in set(exclude_cols)]")

    lines.append(f"norm_cols = out_df[cand_cols].select_dtypes(include={_NUMERIC_DTYPES}).columns.tolist()")
    lines.append(f"{bundle_var}['columns'] = list(norm_cols)")

    lines.append("if not norm_cols:")
    lines.append("    # No numeric columns to normalize; passthrough")
    lines.append(f"    {bundle_var}['stats'] = {{}}")
    lines.append("    pass")
    lines.append("else:")
    lines.append("    # Coerce selected columns to numeric before normalization")
    lines.append("    out_df[norm_cols] = out_df[norm_cols].apply(pd.to_numeric, errors='coerce')")
    if mode == "MINMAX":
        lines.append(f"    _new_min, _new_max = {cfg.new_min}, {cfg.new_max}")
        lines.append("    _span = (_new_max - _new_min)")
        lines.append("    _col_min = out_df[norm_cols].min(axis=0, skipna=True)")
        lines.append("    _col_max = out_df[norm_cols].max(axis=0, skipna=True)")
        lines.append("    stats = {}")
        lines.append("    for col in norm_cols:")
        lines.append("        mn = _col_min.get(col)")
        lines.append("        mx = _col_max.get(col)")
        lines.append("        stats[col] = {")
        lines.append("            'min': None if pd.isna(mn) else float(mn),")
        lines.append("            'max': None if pd.isna(mx) else float(mx),")
        lines.append("        }")
        lines.append(f"    {bundle_var}['stats'] = stats")
        lines.append("    def _minmax_col(s):")
        lines.append("        mn = _col_min.get(s.name)")
        lines.append("        mx = _col_max.get(s.name)")
        lines.append("        rng = (mx - mn) if (mn is not None and mx is not None) else None")
        lines.append("        if rng is None or pd.isna(rng) or rng == 0:")
        lines.append("            # constant/empty column → map to new_min")
        lines.append("            return pd.Series([_new_min] * len(s), index=s.index)")
        lines.append("        return (_new_min + (s - mn) / rng * _span).astype(float)")
        lines.append("    out_df[norm_cols] = out_df[norm_cols].apply(_minmax_col)")
    elif mode == "ZSCORE":
        lines.append("    _col_mean = out_df[norm_cols].mean(axis=0, skipna=True)")
        lines.append("    _col_std  = out_df[norm_cols].std(axis=0, ddof=0, skipna=True)")
        lines.append("    stats = {}")
        lines.append("    for col in norm_cols:")
        lines.append("        mu = _col_mean.get(col)")
        lines.append("        sd = _col_std.get(col)")
        lines.append("        stats[col] = {")
        lines.append("            'mean': None if pd.isna(mu) else float(mu),")
        lines.append("            'std': None if pd.isna(sd) else float(sd),")
        lines.append("        }")
        lines.append(f"    {bundle_var}['stats'] = stats")
        lines.append("    def _zscore_col(s):")
        lines.append("        mu = _col_mean.get(s.name)")
        lines.append("        sd = _col_std.get(s.name)")
        lines.append("        if sd is None or pd.isna(sd) or sd == 0:")
        lines.append("            return pd.Series([0.0] * len(s), index=s.index)")
        lines.append("        return ((s - mu) / sd).astype(float)")
        lines.append("    out_df[norm_cols] = out_df[norm_cols].apply(_zscore_col)")
    elif mode == "DECIMALSCALING":
        lines.append("    _col_absmax = out_df[norm_cols].abs().max(axis=0, skipna=True)")
        lines.append("    stats = {}")
        lines.append("    for col in norm_cols:")
        lines.append("        abs_max = _col_absmax.get(col)")
        lines.append("        if abs_max is None or pd.isna(abs_max):")
        lines.append("            scale = 0")
        lines.append("        else:")
        lines.append("            abs_val = abs(float(abs_max))")
        lines.append("            if abs_val < 1:")
        lines.append("                scale = 0")
        lines.append("            else:")
        lines.append("                scale = int(math.floor(math.log10(abs_val)) + 1)")
        lines.append("        stats[col] = {'scale': scale}")
        lines.append(f"    {bundle_var}['stats'] = stats")
        lines.append("    def _decimal_scale_col(s):")
        lines.append("        abs_max = _col_absmax.get(s.name)")
        lines.append("        if abs_max is None or pd.isna(abs_max):")
        lines.append("            return s.astype(float)")
        lines.append("        abs_val = abs(float(abs_max))")
        lines.append("        if abs_val < 1:")
        lines.append("            scale = 0")
        lines.append("        elif abs_val == 0:")
        lines.append("            scale = 0")
        lines.append("        else:")
        lines.append("            scale = int(math.floor(math.log10(abs_val)) + 1)")
        lines.append("        if scale <= 0:")
        lines.append("            return s.astype(float)")
        lines.append("        factor = float(10 ** scale)")
        lines.append("        return (s / factor).astype(float)")
        lines.append("    out_df[norm_cols] = out_df[norm_cols].apply(_decimal_scale_col)")
    else:
        lines.append(f"    # Unsupported Normalizer mode '{cfg.mode}'; leaving columns unchanged")

    return lines