Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,147 +1,469 @@
|
|
1 |
-
import
|
2 |
-
import random
|
3 |
-
from typing import List, Tuple
|
4 |
|
5 |
-
import
|
|
|
|
|
6 |
import panel as pn
|
7 |
-
from
|
8 |
-
from
|
|
|
9 |
|
10 |
-
pn.extension(
|
|
|
11 |
|
12 |
-
ICON_URLS = {
|
13 |
-
"brand-github": "https://github.com/holoviz/panel",
|
14 |
-
"brand-twitter": "https://twitter.com/Panel_Org",
|
15 |
-
"brand-linkedin": "https://www.linkedin.com/company/panel-org",
|
16 |
-
"message-circle": "https://discourse.holoviz.org/",
|
17 |
-
"brand-discord": "https://discord.gg/AXRHnJU6sP",
|
18 |
-
}
|
19 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
|
|
|
|
27 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
|
37 |
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
|
|
42 |
|
|
|
|
|
|
|
43 |
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
-
|
59 |
-
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
68 |
return
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
|
74 |
-
|
75 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
76 |
return
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
for class_item, class_likelihood in zip(class_items, class_likelihoods):
|
85 |
-
row_label = pn.widgets.StaticText(
|
86 |
-
name=class_item.strip(), value=f"{class_likelihood:.2%}", align="center"
|
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 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from pathlib import Path
|
|
|
|
|
2 |
|
3 |
+
import duckdb
|
4 |
+
import holoviews as hv
|
5 |
+
import pandas as pd
|
6 |
import panel as pn
|
7 |
+
from bokeh.models import HoverTool
|
8 |
+
from langchain.callbacks.base import BaseCallbackHandler
|
9 |
+
from langchain.chat_models import ChatOpenAI
|
10 |
|
11 |
+
pn.extension(sizing_mode="stretch_width", notifications=True)
|
12 |
+
hv.extension("bokeh")
|
13 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
|
15 |
+
RANDOM_NAME_QUERY = """
|
16 |
+
SELECT name, count,
|
17 |
+
CASE
|
18 |
+
WHEN female_percent >= 0.2 AND female_percent <= 0.8 AND male_percent >= 0.2 AND male_percent <= 0.8 THEN 'unisex'
|
19 |
+
WHEN female_percent > 0.6 THEN 'female'
|
20 |
+
WHEN male_percent > 0.6 THEN 'male'
|
21 |
+
END AS gender
|
22 |
+
FROM (
|
23 |
+
SELECT
|
24 |
+
name,
|
25 |
+
MAX(male + female) AS count,
|
26 |
+
(SUM(female) / CAST(SUM(male + female) AS REAL)) AS female_percent,
|
27 |
+
(SUM(male) / CAST(SUM(male + female) AS REAL)) AS male_percent
|
28 |
+
FROM names
|
29 |
+
WHERE name LIKE ?
|
30 |
+
GROUP BY name
|
31 |
+
)
|
32 |
+
WHERE count >= ? AND count <= ?
|
33 |
+
AND gender = ?
|
34 |
+
ORDER BY RANDOM()
|
35 |
+
LIMIT 100
|
36 |
+
"""
|
37 |
|
38 |
+
TOP_NAMES_WILDCARD_QUERY = """
|
39 |
+
SELECT name, SUM(male + female) as count
|
40 |
+
FROM names
|
41 |
+
WHERE lower(name) LIKE ?
|
42 |
+
GROUP BY name
|
43 |
+
ORDER BY count DESC
|
44 |
+
LIMIT 10
|
45 |
+
"""
|
46 |
|
47 |
+
TOP_NAMES_SELECT_QUERY = """
|
48 |
+
SELECT name, SUM(male + female) as count
|
49 |
+
FROM names
|
50 |
+
WHERE lower(name) = ?
|
51 |
+
GROUP BY name
|
52 |
+
ORDER BY count DESC
|
53 |
+
"""
|
54 |
|
55 |
+
DATA_QUERY = """
|
56 |
+
SELECT name, year, male, female, SUM(male + female) AS count
|
57 |
+
FROM names
|
58 |
+
WHERE name in ({placeholders})
|
59 |
+
GROUP BY name, year, male, female
|
60 |
+
ORDER BY name, year
|
61 |
+
"""
|
62 |
|
63 |
|
64 |
+
class StreamHandler(BaseCallbackHandler):
|
65 |
+
def __init__(self, container, initial_text="", target_attr="value"):
|
66 |
+
self.container = container
|
67 |
+
self.text = initial_text
|
68 |
+
self.target_attr = target_attr
|
69 |
|
70 |
+
def on_llm_new_token(self, token: str, **kwargs) -> None:
|
71 |
+
self.text += token
|
72 |
+
setattr(self.container, self.target_attr, self.text)
|
73 |
|
74 |
+
|
75 |
+
class NameChronicles:
|
76 |
+
def __init__(self, refresh=False):
|
77 |
+
super().__init__()
|
78 |
+
self.db_path = Path("names.db")
|
79 |
+
self._initialize_database(refresh=refresh)
|
80 |
+
|
81 |
+
# Main
|
82 |
+
self.holoviews_pane = pn.pane.HoloViews(sizing_mode="stretch_both")
|
83 |
+
self.selection = hv.streams.Selection1D()
|
84 |
+
|
85 |
+
# Sidebar
|
86 |
+
|
87 |
+
# Name Widgets
|
88 |
+
self.names_input = pn.widgets.TextInput(name="Name Input", placeholder="Andrew")
|
89 |
+
self.names_input.param.watch(self._add_name, "value")
|
90 |
+
|
91 |
+
self.names_choice = pn.widgets.MultiChoice(
|
92 |
+
name="Selected Names",
|
93 |
+
options=["Andrew"],
|
94 |
+
solid=False,
|
95 |
+
)
|
96 |
+
self.names_choice.param.watch(self._update_plot, "value")
|
97 |
+
self.names_choice.value = ["Andrew"]
|
98 |
+
|
99 |
+
# Reset Widgets
|
100 |
+
self.clear_button = pn.widgets.Button(
|
101 |
+
name="Clear Names", button_style="outline", button_type="primary"
|
102 |
+
)
|
103 |
+
self.clear_button.on_click(
|
104 |
+
lambda event: setattr(self.names_choice, "value", [])
|
105 |
+
)
|
106 |
+
self.refresh_button = pn.widgets.Button(
|
107 |
+
name="Refresh Plot", button_style="outline", button_type="primary"
|
108 |
+
)
|
109 |
+
self.refresh_button.on_click(self._refresh_plot)
|
110 |
+
|
111 |
+
# Randomize Widgets
|
112 |
+
self.name_pattern = pn.widgets.TextInput(
|
113 |
+
name="Name Pattern", placeholder="*na*"
|
114 |
+
)
|
115 |
+
self.count_range = pn.widgets.IntRangeSlider(
|
116 |
+
name="Peak Count Range",
|
117 |
+
value=(10000, 50000),
|
118 |
+
start=0,
|
119 |
+
end=100000,
|
120 |
+
step=1000,
|
121 |
+
margin=(5, 20),
|
122 |
+
)
|
123 |
+
self.gender_select = pn.widgets.RadioButtonGroup(
|
124 |
+
name="Gender",
|
125 |
+
options=["Female", "Unisex", "Male"],
|
126 |
+
button_style="outline",
|
127 |
+
button_type="primary",
|
128 |
+
)
|
129 |
+
randomize_name = pn.widgets.Button(
|
130 |
+
name="Get Name", button_style="outline", button_type="primary"
|
131 |
+
)
|
132 |
+
randomize_name.param.watch(self._randomize_name, "clicks")
|
133 |
+
self.randomize_pane = pn.Card(
|
134 |
+
self.name_pattern,
|
135 |
+
self.count_range,
|
136 |
+
self.gender_select,
|
137 |
+
randomize_name,
|
138 |
+
title="Get Random Name",
|
139 |
+
collapsed=True,
|
140 |
+
)
|
141 |
+
|
142 |
+
# AI Widgets
|
143 |
+
self.ai_key = pn.widgets.PasswordInput(
|
144 |
+
name="OpenAI Key",
|
145 |
+
placeholder="",
|
146 |
+
)
|
147 |
+
self.ai_prompt = pn.widgets.TextInput(
|
148 |
+
name="AI Prompt",
|
149 |
+
value="Share a little history about the name:",
|
150 |
+
)
|
151 |
+
ai_button = pn.widgets.Button(
|
152 |
+
name="Get Response",
|
153 |
+
button_style="outline",
|
154 |
+
button_type="primary",
|
155 |
+
)
|
156 |
+
ai_button.on_click(self._prompt_ai)
|
157 |
+
self.ai_response = pn.widgets.TextAreaInput(
|
158 |
+
placeholder="",
|
159 |
+
disabled=True,
|
160 |
+
height=350,
|
161 |
+
)
|
162 |
+
self.ai_pane = pn.Card(
|
163 |
+
self.ai_key,
|
164 |
+
self.ai_prompt,
|
165 |
+
ai_button,
|
166 |
+
self.ai_response,
|
167 |
+
collapsed=True,
|
168 |
+
title="Ask AI",
|
169 |
+
)
|
170 |
+
|
171 |
+
# Database Methods
|
172 |
+
|
173 |
+
def _connect_database(self):
|
174 |
+
"""
|
175 |
+
Connect to the database.
|
176 |
+
"""
|
177 |
+
return duckdb.connect(database=str(self.db_path))
|
178 |
+
|
179 |
+
def _initialize_database(self, refresh):
|
180 |
+
"""
|
181 |
+
Initialize database with data from the Social Security Administration.
|
182 |
+
"""
|
183 |
+
if not refresh and self.db_path.exists():
|
184 |
return
|
185 |
+
|
186 |
+
df = pd.concat(
|
187 |
+
[
|
188 |
+
pd.read_csv(
|
189 |
+
path,
|
190 |
+
header=None,
|
191 |
+
names=["state", "gender", "year", "name", "count"],
|
192 |
+
)
|
193 |
+
for path in Path("data").glob("*.TXT")
|
194 |
+
]
|
195 |
+
)
|
196 |
+
df_processed = (
|
197 |
+
df.groupby(["gender", "year", "name"], as_index=False)[["count"]]
|
198 |
+
.sum()
|
199 |
+
.pivot(index=["name", "year"], columns="gender", values="count")
|
200 |
+
.reset_index()
|
201 |
+
.rename(columns={"F": "female", "M": "male"})
|
202 |
+
.fillna(0)
|
203 |
+
)
|
204 |
+
with self._connect_database() as conn:
|
205 |
+
conn.execute("DROP TABLE IF EXISTS names")
|
206 |
+
conn.execute("CREATE TABLE names AS SELECT * FROM df_processed")
|
207 |
+
|
208 |
+
def _query_names(self, names):
|
209 |
+
"""
|
210 |
+
Query the database for the given name.
|
211 |
+
"""
|
212 |
+
dfs = []
|
213 |
+
for name in names:
|
214 |
+
if "*" in name or "%" in name:
|
215 |
+
name = name.replace("*", "%")
|
216 |
+
top_names_query = TOP_NAMES_WILDCARD_QUERY
|
217 |
+
else:
|
218 |
+
top_names_query = TOP_NAMES_SELECT_QUERY
|
219 |
+
with self._connect_database() as conn:
|
220 |
+
top_names = (
|
221 |
+
conn.execute(top_names_query, [name.lower()])
|
222 |
+
.fetch_df()["name"]
|
223 |
+
.tolist()
|
224 |
+
)
|
225 |
+
if len(top_names) == 0:
|
226 |
+
pn.state.notifications.info(f"No names found matching {name!r}")
|
227 |
+
continue
|
228 |
+
data_query = DATA_QUERY.format(
|
229 |
+
placeholders=", ".join(["?"] * len(top_names))
|
230 |
+
)
|
231 |
+
df = conn.execute(data_query, top_names).fetch_df()
|
232 |
+
dfs.append(df)
|
233 |
+
|
234 |
+
if len(dfs) > 0:
|
235 |
+
self.df = pd.concat(dfs).drop_duplicates(
|
236 |
+
subset=["name", "year", "male", "female"]
|
237 |
+
)
|
238 |
+
else:
|
239 |
+
self.df = pd.DataFrame(columns=["name", "year", "male", "female"])
|
240 |
+
|
241 |
+
# Widget Methods
|
242 |
+
|
243 |
+
def _randomize_name(self, event):
|
244 |
+
with self._connect_database() as conn:
|
245 |
+
name_pattern = self.name_pattern.value.lower()
|
246 |
+
if not name_pattern:
|
247 |
+
name_pattern = "%"
|
248 |
+
else:
|
249 |
+
name_pattern = name_pattern.replace("*", "%")
|
250 |
+
count_range = self.count_range.value
|
251 |
+
gender_select = self.gender_select.value.lower()
|
252 |
+
random_names = (
|
253 |
+
conn.execute(
|
254 |
+
RANDOM_NAME_QUERY, [name_pattern, *count_range, gender_select]
|
255 |
+
)
|
256 |
+
.fetch_df()["name"]
|
257 |
+
.tolist()
|
258 |
+
)
|
259 |
+
if random_names:
|
260 |
+
for i in range(len(random_names)):
|
261 |
+
random_name = random_names[i]
|
262 |
+
if random_name in self.names_choice.value:
|
263 |
+
continue
|
264 |
+
self.names_input.value = random_name
|
265 |
+
break
|
266 |
+
else:
|
267 |
+
pn.state.notifications.info(
|
268 |
+
"All names matching the criteria are already added!"
|
269 |
+
)
|
270 |
+
else:
|
271 |
+
pn.state.notifications.info("No names found matching the criteria!")
|
272 |
+
|
273 |
+
def _add_name(self, event):
|
274 |
+
name = event.new.strip().title()
|
275 |
+
self.names_input.value = ""
|
276 |
+
if not name:
|
277 |
return
|
278 |
+
elif name in self.names_choice.options and name in self.names_choice.value:
|
279 |
+
pn.state.notifications.info(f"{name!r} already added!")
|
280 |
+
return
|
281 |
+
elif len(self.names_choice.value) > 10:
|
282 |
+
pn.state.notifications.info(
|
283 |
+
"Maximum of 10 names allowed; please remove some first!"
|
|
|
|
|
|
|
|
|
284 |
)
|
285 |
+
return
|
286 |
+
value = self.names_choice.value.copy()
|
287 |
+
options = self.names_choice.options.copy()
|
288 |
+
if name not in options:
|
289 |
+
options.append(name)
|
290 |
+
if name not in value:
|
291 |
+
value.append(name)
|
292 |
+
self.names_choice.param.update(
|
293 |
+
options=options,
|
294 |
+
value=value,
|
295 |
+
)
|
296 |
+
|
297 |
+
def _prompt_ai(self, event):
|
298 |
+
if not self.ai_key.value:
|
299 |
+
pn.state.notifications.info("Please enter an API key!")
|
300 |
+
return
|
301 |
+
|
302 |
+
if not self.ai_prompt.value:
|
303 |
+
pn.state.notifications.info("Please enter a prompt!")
|
304 |
+
return
|
305 |
+
|
306 |
+
stream_handler = StreamHandler(self.ai_response)
|
307 |
+
chat = ChatOpenAI(
|
308 |
+
max_tokens=500,
|
309 |
+
openai_api_key=self.ai_key.value,
|
310 |
+
streaming=True,
|
311 |
+
callbacks=[stream_handler],
|
312 |
+
)
|
313 |
+
self.ai_response.loading = True
|
314 |
+
try:
|
315 |
+
if self.selection.index:
|
316 |
+
names = [self._name_indices[self.selection.index[0]]]
|
317 |
+
else:
|
318 |
+
names = self.names_choice.value[:3]
|
319 |
+
chat.predict(f"{self.ai_prompt.value} {names}")
|
320 |
+
finally:
|
321 |
+
self.ai_response.loading = False
|
322 |
+
|
323 |
+
# Plot Methods
|
324 |
+
|
325 |
+
def _click_plot(self, index):
|
326 |
+
gender_nd_overlay = hv.NdOverlay(kdims=["Gender"])
|
327 |
+
if not index:
|
328 |
+
return hv.NdOverlay(
|
329 |
+
{
|
330 |
+
"curve": self._curve_nd_overlay,
|
331 |
+
"scatter": self._scatter_nd_overlay,
|
332 |
+
"label": self._label_nd_overlay,
|
333 |
+
}
|
334 |
)
|
335 |
+
|
336 |
+
name = self._name_indices[index[0]]
|
337 |
+
df_name = self.df.loc[self.df["name"] == name].copy()
|
338 |
+
df_name["female"] += df_name["male"]
|
339 |
+
gender_nd_overlay["Male"] = hv.Area(
|
340 |
+
df_name, ["year"], ["male"], label="Male"
|
341 |
+
).opts(alpha=0.3, color="#add8e6", line_alpha=0)
|
342 |
+
gender_nd_overlay["Female"] = hv.Area(
|
343 |
+
df_name, ["year"], ["male", "female"], label="Female"
|
344 |
+
).opts(alpha=0.3, color="#ffb6c1", line_alpha=0)
|
345 |
+
return hv.NdOverlay(
|
346 |
+
{
|
347 |
+
"curve": self._curve_nd_overlay[[index[0]]],
|
348 |
+
"scatter": self._scatter_nd_overlay,
|
349 |
+
"label": self._label_nd_overlay[[index[0]]].opts(text_color="black"),
|
350 |
+
"gender": gender_nd_overlay,
|
351 |
+
},
|
352 |
+
kdims=["Gender"],
|
353 |
+
).opts(legend_position="top_left")
|
354 |
+
|
355 |
+
@staticmethod
|
356 |
+
def _format_y(value):
|
357 |
+
return f"{value / 1000}k"
|
358 |
+
|
359 |
+
def _update_plot(self, event):
|
360 |
+
names = event.new
|
361 |
+
print(names)
|
362 |
+
self._query_names(names)
|
363 |
+
|
364 |
+
self._scatter_nd_overlay = hv.NdOverlay()
|
365 |
+
self._curve_nd_overlay = hv.NdOverlay(kdims=["Name"]).opts(
|
366 |
+
gridstyle={"xgrid_line_width": 0},
|
367 |
+
show_grid=True,
|
368 |
+
fontscale=1.28,
|
369 |
+
xlabel="Year",
|
370 |
+
ylabel="Count",
|
371 |
+
yformatter=self._format_y,
|
372 |
+
legend_limit=0,
|
373 |
+
padding=(0.2, 0.05),
|
374 |
+
title="Name Chronicles",
|
375 |
+
responsive=True,
|
376 |
+
)
|
377 |
+
self._label_nd_overlay = hv.NdOverlay(kdims=["Name"])
|
378 |
+
hover_tool = HoverTool(
|
379 |
+
tooltips=[("Name", "@name"), ("Year", "@year"), ("Count", "@count")],
|
380 |
+
)
|
381 |
+
self._name_indices = {}
|
382 |
+
scatter_cycle = hv.Cycle("Category10")
|
383 |
+
curve_cycle = hv.Cycle("Category10")
|
384 |
+
label_cycle = hv.Cycle("Category10")
|
385 |
+
for i, (name, df_name) in enumerate(self.df.groupby("name")):
|
386 |
+
df_name_total = df_name.groupby(
|
387 |
+
["name", "year", "male", "female"], as_index=False
|
388 |
+
)["count"].sum()
|
389 |
+
df_name_total["male"] = df_name_total["male"] / df_name_total["count"]
|
390 |
+
df_name_total["female"] = df_name_total["female"] / df_name_total["count"]
|
391 |
+
df_name_peak = df_name.loc[[df_name["count"].idxmax()]]
|
392 |
+
df_name_peak[
|
393 |
+
"label"
|
394 |
+
] = f'{df_name_peak["name"].item()} ({df_name_peak["year"].item()})'
|
395 |
+
|
396 |
+
hover_tool = HoverTool(
|
397 |
+
tooltips=[
|
398 |
+
("Name", "@name"),
|
399 |
+
("Year", "@year"),
|
400 |
+
("Count", "@count{(0a)}"),
|
401 |
+
("Male", "@male{(0%)}"),
|
402 |
+
("Female", "@female{(0%)}"),
|
403 |
+
],
|
404 |
+
)
|
405 |
+
self._scatter_nd_overlay[i] = hv.Scatter(
|
406 |
+
df_name_total, ["year"], ["count", "male", "female", "name"], label=name
|
407 |
+
).opts(
|
408 |
+
color=scatter_cycle,
|
409 |
+
size=4,
|
410 |
+
alpha=0.15,
|
411 |
+
marker="y",
|
412 |
+
tools=["tap", hover_tool],
|
413 |
+
line_width=3,
|
414 |
+
show_legend=False,
|
415 |
+
)
|
416 |
+
self._curve_nd_overlay[i] = hv.Curve(
|
417 |
+
df_name_total, ["year"], ["count"], label=name
|
418 |
+
).opts(
|
419 |
+
color=curve_cycle,
|
420 |
+
tools=["tap"],
|
421 |
+
line_width=3,
|
422 |
+
)
|
423 |
+
self._label_nd_overlay[i] = hv.Labels(
|
424 |
+
df_name_peak, ["year", "count"], ["label"], label=name
|
425 |
+
).opts(
|
426 |
+
text_align="right",
|
427 |
+
text_baseline="bottom",
|
428 |
+
text_color=label_cycle,
|
429 |
+
)
|
430 |
+
self._name_indices[i] = name
|
431 |
+
self.selection.source = self._curve_nd_overlay
|
432 |
+
if len(self._name_indices) == 1:
|
433 |
+
self.selection.update(index=[0])
|
434 |
+
else:
|
435 |
+
self.selection.update(index=[])
|
436 |
+
self.dynamic_map = hv.DynamicMap(
|
437 |
+
self._click_plot, kdims=[], streams=[self.selection]
|
438 |
+
).opts(responsive=True)
|
439 |
+
self._refresh_plot()
|
440 |
+
|
441 |
+
def _refresh_plot(self, event=None):
|
442 |
+
self.holoviews_pane.object = self.dynamic_map.clone()
|
443 |
+
|
444 |
+
def view(self):
|
445 |
+
reset_row = pn.Row(self.clear_button, self.refresh_button)
|
446 |
+
data_url = pn.pane.Markdown(
|
447 |
+
"<center>Data from the <a href='https://www.ssa.gov/oact/babynames/limits.html' "
|
448 |
+
"target='_blank'>U.S. Social Security Administration</a></center>",
|
449 |
+
align="end",
|
450 |
+
)
|
451 |
+
sidebar = pn.Column(
|
452 |
+
self.names_input,
|
453 |
+
self.names_choice,
|
454 |
+
reset_row,
|
455 |
+
pn.layout.Divider(),
|
456 |
+
self.randomize_pane,
|
457 |
+
self.ai_pane,
|
458 |
+
data_url,
|
459 |
+
)
|
460 |
+
template = pn.template.FastListTemplate(
|
461 |
+
sidebar=[sidebar],
|
462 |
+
main=[self.holoviews_pane],
|
463 |
+
title="Name Chronicles",
|
464 |
+
theme="dark",
|
465 |
+
)
|
466 |
+
return template
|
467 |
+
|
468 |
+
|
469 |
+
NameChronicles().view().servable()
|