Archie
@archieemwood
Followers
1,097
Following
190
Media
390
Statuses
1,325
the data driven dad (TM) 💻: @evidence_dev
Joined March 2015
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
TLP HAIRPIN OF LOVE
• 480922 Tweets
#ปิ่นภักดิ์EP5
• 467143 Tweets
Charles
• 257030 Tweets
#ItalianGP
• 256726 Tweets
نيوكاسل
• 197266 Tweets
Monza
• 161180 Tweets
Ferrari
• 153611 Tweets
India Will Become A Vishwaguru
• 135077 Tweets
Liverpool
• 131577 Tweets
Oscar
• 90737 Tweets
DAOU IN PATTAYA
• 81239 Tweets
Lando
• 66536 Tweets
Norris
• 63660 Tweets
McLaren
• 58318 Tweets
Colapinto
• 55818 Tweets
#MUNLIV
• 41318 Tweets
Piastri
• 41152 Tweets
Spurs
• 36431 Tweets
プロジェクトKV
• 32308 Tweets
台風11号
• 31590 Tweets
Newcastle
• 31445 Tweets
Sainz
• 29670 Tweets
ブルアカ
• 28209 Tweets
Tottenham
• 27272 Tweets
#f1jp
• 26972 Tweets
aiko
• 23939 Tweets
渡海先生
• 19402 Tweets
Nicolas Jackson
• 18411 Tweets
天城先生
• 15667 Tweets
Amad
• 14292 Tweets
wrwrd
• 13704 Tweets
Isak
• 11149 Tweets
one of the more awesome things you can do with
@duckdb
query geospatial data straight from github
this is from the excellent Natural Earth Data, and contains the boundaries for all countries in the world
4
43
413
When I was a Chief of Staff, one meeting generated more stress than *all the others* in the week.
What was it?
Our Tuesday AM 📊Weekly Business Review📊
(1/x)
10
25
297
okay so how about this
@duckdb
's statistical functions have enough coverage that you can basically recreate facebook prophet in SQL
time series forecasting in SQL, here we come
4
23
268
it can be tricky to do forecasting in SQL, because SQL isn't really designed for operating on data that doesn't exist
in
@duckdb
it's simple, by combining generate_series() with unnest()
2
20
236
awesome
@duckdb
ad
oh wait, was it for something else?
Why all data scientists should learn Polars in Python.
This is why: 🧵
7
92
521
7
13
216
69yo mother in law asked about my job:
me “I work in BI”
her: “is that like AI?”
me: “confusingly not”
didnt have the heart to tell her i also do CI
6
17
206
to build this whole app, I wrote <60 lines of code
~ 20 lines of SQL
+ 35 lines of Markdown
2
7
161
okay so why is this such a cool
@duckdb
feature?
it means you group by multiple different dimensions in one query!
3
12
136
when you run a query in
@duckdb
, it outputs results in a nice box
but there are 20 other **really useful** formats you can use with .mode
here are 7 of my favourite
.mode duckbox (default)
0
0
9
4
14
132
also had some fun investigating command line plotting tools for the same dataset
this is Youplot by Red Data Tools
2
9
114
part of what makes the SQL so terse is that it's
@duckdb
`group by all` is such a quality of life feature
to build this whole app, I wrote <60 lines of code
~ 20 lines of SQL
+ 35 lines of Markdown
2
7
161
0
6
102
change my mind:
all radar charts should just be another kind of chart
23
2
97
exploring
@duckdb
's recent performance data with
@evidence_dev
is such a delight with the new SQL autocomplete
3
7
88
and since
@evidence_dev
uses DuckDB, you can get geospatial data directly into Evidence like this
one of the more awesome things you can do with
@duckdb
query geospatial data straight from github
this is from the excellent Natural Earth Data, and contains the boundaries for all countries in the world
4
43
413
3
9
86
underrated feature of
@evidence_dev
it creates a parquet catalog from your data sources and publishes them with the app
so you’re not waiting on the data warehouse for your dashboard to load
2
9
86
just released
@duckdb
VSCode SQLTools v0.0.5, with support for DuckDB v0.9.1
- database explorer, query runner, autocompletion
- connect to a local, in memory or
@motherduck
DB
3
9
86
NLP exploration with
@duckdb
...
DuckDB doesn't have libraries like nltk for NLP, but that doesn't mean you can't do NLP, you just have to implement it from scratch
6
4
84
just recorded some new getting started videos on how to use
@evidence_dev
1 / Installing Evidence from VS Code
2
8
82
this is really some cool innovation
it uses
@duckdb
webassembly to reissue SQL queries in the browser against data files
this means you can explore very large datasets, with extremely low latency
Evidence Universal SQL is coming!
This brings multi data source support, and fast interactivity to Evidence
Sign up for the community preview next week here:
3
11
68
1
9
81
continuing on my vscode extension tour, just published an open source extension for duckdb
supports connecting to local, in-memory, and
@motherduck
instances
0
16
80
brb query running
1
1
13
3
7
77
and if you can get it into duckdb, then you can get it in
@evidence_dev
so i replicated the twitter stats portal in Evidence
X premium users get analytics on how much people are viewing their content
if you open the devtools network tab, you can see all that data gets delivered in a JSON blob
it's a pretty hairy nested JSON, but as always, only a query away from a nice table in
@duckdb
6
4
66
1
5
76
Update, I now have all the (roughly 100) coalesce session videos online - so you can watch any you missed!
All indexed and easy to search and find
2
12
75
just shipped read-write support in SQLTools for DuckDB
edit your local
@duckdb
or
@motherduck
, and see it update, all in vscode
0
8
71
in
@evidence_dev
you can annotate notable events in your database
and then display them on your charts
all in SQL
3
2
69
writing a
@duckdb
extension for the first time
here's a quick guide
DuckDB extensions allow you to add:
- custom functions (scalar, aggregate, table)
- custom data types
- custom output formats
4
8
67
X premium users get analytics on how much people are viewing their content
if you open the devtools network tab, you can see all that data gets delivered in a JSON blob
it's a pretty hairy nested JSON, but as always, only a query away from a nice table in
@duckdb
6
4
66
most charting tools require inordinate amounts of config for a decent looking chart
in
@evidence_dev
a great looking chart is 5 lines of code
3
6
64
out here doing my best to shill
@duckdb
and i have to read stuff like this
guys you gotta shill this harder, did you know you can create a duckdb database from a csv and then just query it from dbeaver? Very very useful:
9
6
123
1
1
62
@PradyuPrasad
so if I look after my friends child for four hours and they pay me $100 and then tomorrow they look after my child and pay me $100 for 4 hrs, that increases GDP by $200
But if we just agree to swap, then it wouldn’t
There are lots of uncounted exchanges of services
0
0
60
this is all SQL for the dynamic forecasting
@evidence_dev
for the plotting
okay so how about this
@duckdb
's statistical functions have enough coverage that you can basically recreate facebook prophet in SQL
time series forecasting in SQL, here we come
4
23
268
4
3
61
how I feel when I rip open
@duckdb
Psilocybin leads to a profound shift in brain connectivity patterns.
🧵1/10
358
2K
19K
1
3
60
X also stores a JSON blob with the stats for each of your individual posts, so i've made a page for each post
next stop: what makes a hit tweet?
and if you can get it into duckdb, then you can get it in
@evidence_dev
so i replicated the twitter stats portal in Evidence
1
5
76
1
6
60
this is a use case I really like small multiples for
exploring
@duckdb
's recent performance data with
@evidence_dev
is such a delight with the new SQL autocomplete
3
7
88
0
6
55
made an incredibly minimal CLI plotting tool
allows you to plot using csv or command line arguments
what should I add next?
4
0
52
next step: turn the text into a "Bag of Words" representation
it's a sparse table with one row per tweet, one column per word, count of words
in most SQL dialects this would be impossible - you need to know the columns ahead of time to pivot
not with
@duckdb
's PIVOT syntax
one of the more interesting parts of this is you need a list of "stop words" - common words that aren't interesting when doing text analysis:
I, you, they, about, in, that
they aren't built into duckdb, but there are lots of great public lists, so just a SELECT away
1
3
43
2
3
51
one of the more interesting parts of this is you need a list of "stop words" - common words that aren't interesting when doing text analysis:
I, you, they, about, in, that
they aren't built into duckdb, but there are lots of great public lists, so just a SELECT away
NLP exploration with
@duckdb
...
DuckDB doesn't have libraries like nltk for NLP, but that doesn't mean you can't do NLP, you just have to implement it from scratch
6
4
84
1
3
43
just using
@evidence_dev
's Dimension Grid for the first time, and it is a really nice way to understand your data
0
4
43
US BLS publishes avg salaries for 1,100 jobs at a state / metro level
6M rows per year = child's play for
@duckdb
data scientist salaries by state, 2023
3
8
42
now I get to combine my three favourite things
Data + Dad + Dashboards
6
2
42
in most bi tools it’s hard to create layered maps
because drag and drop UIs for creating maps layers suck
in
@evidence_dev
you compose maps from layers with code
based on the wonderful, open source
@LeafletJS
3
8
41
for the
@evidence_dev
hackathon at our offsite this year, I built evidence-new
drop your CSV into a new Evidence app in 20 seconds
4
2
38
who needs stats packages when you have SQL
0
5
37
if you listen carefully, you can hear the sound of no-one reading your documentation
2
3
35
one of the low-key best things about
@evidence_dev
is how terse the charting syntax is
6 lines of code -> beautiful viz
1
4
34
have been waiting until someone grabbed the olympics subdomain on
@evidence_dev
was not disappointed
3
6
33
stuff coming out from
@evidence_dev
today:
1) a really sweet dimension grid component - kudos to
@AdamMcaskill
for shipping this one
1
5
33
so an absolutely mega set of features released today for
@evidence_dev
here are some of my favourites 👇
3
4
30
with
@evidence_dev
in the top 10 in product hunt today, I wanted to share 10 awesome projects the community has built 🧵
2
12
29
if you missed any of the talks at dbt coalesce this year, I've published an "unofficial coalesce app" where you can find slides for ~30% of the sessions.
planning to add links to youtube vids after they are published too
1
1
27
One of the other cool things that is supported in universal SQL is non database data sources like... google sheets
Evidence Universal SQL is coming!
This brings multi data source support, and fast interactivity to Evidence
Sign up for the community preview next week here:
3
11
68
3
4
26
90% of data team-biz stakeholder interactions
Data: Hey, we're from the data team, and we're here to help. If there's a problem that we can solve with data - ask!
Biz: I have this excel, it's taken me 100 hours to build, and takes me 5 more each week. Can you automate?
Data:💀
2
1
26
advent of code in SQL
day 1 part deux
this regex feels pretty dirty - anyone got anything better?
4
0
26
0
0
25
Did a demo call with
@Oracle
MySQL Heatwave today, and must say the pitch is good:
1. One database for both transactional and analytical queries: Less ETL needed.
2. It’s 3x faster and 5-8x cheaper than [Snowflake / Aurora / BigQuery]
3. Support for querying CSV, parquet files
4
2
25
this is a cool feature
let me look up my job title in the BLS occupation list to see what bucket I'm closest to
results seem good
Excited to share that
@motherduck
now makes computing semantic similarities in
@duckdb
as simple as a one-liner!
2
6
64
1
2
23
rendering a new component on the fly with
@evidence_dev
(very directed output for now)
@archieemwood
@evidence_dev
Rendering a new component on the fly would be sick, but could it work reliably enough??
0
0
1
3
1
23
Just released DuckDB 0.10.0 support for the VSCode SQLTools extension
Thanks
@_ccer0z
for the PR to get it over the line!
1
1
22
v excited for this
Development builds of
@DuckDB
have a version of histogram() () that lets you specify the bins. () So now we can do stuff like:
1
3
20
1
0
20
making your own interactive components is super easy in Evidence now
made this "freetext search in a chart" in a few mins
cool: 💯
useful: 🤔
1
1
20
@matsonj
continuing to build the coolest stuff with
@evidence_dev
a sweet "what if" matchup simulator between historical NBA teams
1
2
20
@FrankBoerman
@duckdb
duckdb>
install spatial;
load spatial;
select * from '
https://whatever/shapefile/on/the/internet.geojson';
1
3
18
@evidence_dev
60 lines of code in
@evidence_dev
here's the source code if you want to run it on your own account
2
2
18
tables containing lots of KPIs in rows are really common, but somehow they are really hard in most BI tools
messing around with the UX in
@evidence_dev
thoughts?
3
1
18
Just arrived for our wedding in Lyon and
@AirCanada
has lost our bag, with my fiancée’s wedding dress in it.
Can you help get the dress to us before Friday and save the wedding? +1 647 207 7690
Flew AC7968, AC867
Have willing friends in MTL, NYC & Paris so can get creative!
5
10
18
my first database was as csv file committed to git
it worked but made my commit history look kinda crazy
I just read a 127 comment thread about people arguing whether or not Excel was a database or not
262
102
2K
1
1
17