has anyone built a nice UI tool for interacting with datascript?

something where (1) I can browse all elements, (2) write queries, and (3) run transactions

basically I want a "datascript shell" as an HTML UI element

I’m not aware of anything

but I thought about something like that, of course :)

@tonsky: we should sign a petition to get the dev of datascript to implement it 🙂

yeah, I know, that guy is so lazy!

@tonsky: have you played with datastore? it doesn't support full datascript, like join, but the two seems very closely matched

which datastore is that?

i always want to build an "adminer" kind of thing, but i never have the time

also, what's the simplest setup for toggling a value in datascript between true/false?

@tonsky: Google Cloud https://cloud.google.com/appengine/docs/standard/python/datastore/ It's basically an EAV store, where Attributes are called "Property"

iirc, the indexing model are also very siilar

Query q =
    new Query("Person")
        .setFilter(new FilterPredicate("lastName", FilterOperator.EQUAL, targetLastName));

PreparedQuery pq = datastore.prepare(q);
Entity result = pq.asSingleEntity();

@misha: yeah, wouldn't you rather query that using datalog?

@misha: help me convince @tonsky to build a datascript layer over google data store 🙂

build it yourself

I'm in the process of doing exactly that

as my backend is gae, so I have no choice but to use datastore

how building over datastore is useful? what is the usecase?

anyone a suggestion for boolean toggling?

focus on the short term pls 🙂

"as my backend is gae, so I have no choice but to use datastore make people implement things for me"

@misha: I'm not sure if I'm missing your humor, or if you're missing my humor, but the 🙂 in "@misha: help me convince @tonsky to build a datascript layer over google data store :-)" meant I wasn't really expecting tonsky to implement it

how did slack turn into twitchchat

@ashnur https://github.com/tonsky/datascript/blob/master/test/datascript/test/transact.cljc#L165-L191

but I do think, from a tech perpsective, gae datastore is an excellent persistenticy backend for datascript

there is also :db.fn/cas, but writing toggle fn would be cleaner(?) in use I guess

cas is not available for me

tried that yesterday

js? tough luck 😄

datascript-mori module

and i am not sure which part of this test is the response to my question, seems parallel

no idea what that is, but see if you can write a transact function

you want to toggle w/o knowing current value, right?

would be the most convenient

write a transact function. link is a tx-fn example in a test

otherwise, if you already hold value in the code:

(ds/transact! conn [[:db/add id :my.bool/flag (not v)]])

if i do that it will fill up the db, adding a new value every time, wouldn't it?

i had that issue before

seems like for most things i should use db functions

1. :my.bool/flag has to be {:db/cardinality :db.cardinality/one} to be updated "in-place" 2. datascript does not keep old values like datomic does, so if you don't keep db-before-transaction db value yourself - it will be updated "in-place"

no idea what datascript-mori does

(let [conn (ds/create-conn {:my.bool/flag {:db/cardinality :db.cardinality/one}})]
  (ds/transact! conn [{:my.bool/flag true}])
  (prn [:bfore (into [] (ds/datoms @conn :eavt))])
  (ds/transact! conn [[:db/add 1 :my.bool/flag false]])
  (prn [:after (into [] (ds/datoms @conn :eavt))]))

;=>
[:bfore [#datascript/Datom [1 :my.bool/flag true 536870913 true]]]
[:after [#datascript/Datom [1 :my.bool/flag false 536870914 true]]]

datascript-mori is just convenience module for js folks so we can use mori functions with datascript entities

because clojurescript is so good, if you don't compile stuff together, they don't work

anyway, if it does not hold on to pre-transaction db value (version) for you - datoms count will not grow after cardinality/one attribute change

@ashnur

@misha thanks, it was the cardinality/one thing that i forgot about

is this possible for nested stuff too? so when i save an object with reference, i would like that reference to not add new stuff but to replace the old ones. (right now i am saving address path from history module and each location change adds 3 new datoms)

problem is that the same attribute might not be like this for a different ident

so i am not sure i want to add to every attribute in location object this cardinality/one thing because then i can't save more stuff with same attribute name under different entities

in fact, i tried but obvious it doesn't work because this is a different issue, even though i thought it's the same 😞

right now i am not even sure if it's possible to do this without retracting explicitly 😞

oh, it works, i just had to set everything unique/identity :((

I have an impression you are confusing cardinality with identity

;; :foo :cardinality/one is
{:foo "a"}
;; :foo :cardinality/many is
{:foo ["a"]}

that would be interesting. cardinality is about counting stuff and identity is about matching stuff

then I did not quite understand you question about growing datoms count on nested ref change

https://gist.github.com/ashnur/c7d8a6721ce7440a1df42f90dc572120

i really wish i could avoid having to add pathname search and hash to the schema

especially with "identity"

db = ds.db_with(db, vector( hashMap(DB_ID, vector("DB_IDENT", "history") , "current", toClj(removeNull(history.location)))))

i don't speak twitch language

js looks ugly

yeah, i am not going to argue about the aesthetics of the language if it's ok, i would much rather just find out if i am doing something wrong

having hard time actually understanding what that code represents

which part of it is hard?

vector is just cljs vector

DB_ID is the :db/id, removeNull removes all null values from any object.

i am using history module and trying to save the current location object into datascript

and i would like to only have the last one, but in order for that to work, i have to add each of the attributes as "identity"

so you want to put current url into datascript as a parsed map of path, search, hash, etc. right?

and want to have only 1 such destructured url in an entire db, correct?

yes

history.location is {pathname, hash, search}

(and some null stuff that is removed

(let [conn (ds/create-conn)
      temp-id -1
      report (ds/transact! conn [{:db/id temp-id :url/path "foo" :url/hash "bar"}])
      actual-id (get (:tempids report) temp-id)]
  (prn (into [] (ds/datoms @conn :eavt)))
  (ds/transact! conn [[:db/add actual-id :url/hash "baz"]])
  (prn (into [] (ds/datoms @conn :eavt))))

=>
[#datascript/Datom [1 :url/hash "bar" 536870913 true]
 <#C07V8N22C|datascript>/Datom [1 :url/path "foo" 536870913 true]]
[#datascript/Datom [1 :url/hash "baz" 536870914 true]
 <#C07V8N22C|datascript>/Datom [1 :url/path "foo" 536870913 true]]

that updates in parsed url in place

no schema needed at all

ok, so i also have to be able to find it later

wait, i can't read code, sry

let me try again ;D

basically all you need to do, is to use the same id when you update stuff

ident works too, yeah, why not

dunno, seems to me you just took 2 random attributes and saved some data into it

what am i missing here

(let [conn (ds/create-conn {:db/ident {:db/unique :db.unique/identity}})
      report (ds/transact! conn [{:db/ident "history" :url/path "foo" :url/hash "bar"}])]
  (prn (into [] (ds/datoms @conn :eavt)))
  (ds/transact! conn [[:db/add [:db/ident "history"] :url/hash "baz"]])
  (prn (into [] (ds/datoms @conn :eavt))))

=>
[#datascript/Datom [1 :db/ident "history" 536870913 true]
 <#C07V8N22C|datascript>/Datom [1 :url/hash "bar" 536870913 true]
 <#C07V8N22C|datascript>/Datom [1 :url/path "foo" 536870913 true]]
[#datascript/Datom [1 :db/ident "history" 536870913 true]
 <#C07V8N22C|datascript>/Datom [1 :url/hash "baz" 536870914 true]
 <#C07V8N22C|datascript>/Datom [1 :url/path "foo" 536870913 true]]

yeah, so what i ahve more is the "current" attribute with ":db/valueType": ":db.type/ref",

3 datoms before and after value change

so, first of all, thanks for your help, i really do not want to seem as ungrateful, this is definitely a good solution for what i described

however i am afraid this means that i can't have any kind of functions working over datascript because i can't use attributes like "current"

i mean general functions.

my vision of all this was that if i have a "current" with reference then i can just save into that and then have a function that takes just the ident name and gives back the current value

if for each value i have to have a different attribute

not sure if it's worth it

can you paste a json, representing what you want to have in a db? I think we are running circles here

1) you want to be able to get current url from db 2) and to keep all previous ones in db too ?

i don't have a json, but i have this: https://gist.github.com/ashnur/44b5535466629d0f01b1d85e97a496c4

no, i don't want to keep previous ones.

you can see that i have "current" both for "history" and for "myAddr"

(and if i had saved i could have it for "contraAddr" too

function getCurrentAttr(ident, attr){
    return toJs(ds.q(Datalog.Q`[:find ?v . :in $ ?e ?a :where [?e "current" ?r][?r ?a ?v]]`, db, vector("DB_IDENT", ident), attr))
}

you want attribute named "current" on different entities of different type?

yeah

and you think you can't have those because?

[[:db/add [:db/ident :myAddr] :current "foo"]]

however both "current" attributes has to have the same schema

to avoid this you need to use namespaces. in clojure it would look just like this:

:address/current ...
:history/current ...

how it would look like in js - have no idea, other than "current_history" and "current_address"

this is datascript

datascript is javascript

🙂

anyway, thanks for pointing this out, i always run into stuff like this, expecting stuff to work when it can't

I mean "conveniently" for you, so you could pull data from datascript and use it with no extra hustle

if you have the same schema for the "current" attribute, then there is no problem at all

and I still don't fully understand the problem

that looks like XBOCT

@misha so when i used schema that i shared with you but without pathname, search, hash specified as identity, and I added new location, the current got updated as it should but instead of just replacing the old pathname,search,hash, it added them with new entity id.

https://gist.github.com/ashnur/db533c6aa47f933b39b8867c2352e514 like this

what exactly did you transact?

that's the code i've shown to you before:

db = ds.db_with(db, vector( hashMap(DB_ID, vector("DB_IDENT", "history") , "current", toClj(removeNull(location)))))

what is in location? hash-map?

and if you would have only 1 entity url in db - what would current represent/contain?

as i described before, it is a js object with pathname, search, hash attributes and some others that get removed because they have null values

what is entity url?

yeah, that object

you need to set an id inside that object

modifying javascript objects is a big antipattern

if you omit id - datascript treats it as a new entity coming in

i see

well, i should be able to use conj

i just don't know the id, that's why i use idents

if you transact:

{:db/ident "history"
 :current {:foo bar}}

to update existing you need to transact

{:db/ident "history"
 :current {:db/ident "history" :foo bar}}

or whatever db/ident :foo belongs to

then, I just don't understand why would you have a "current" attribute on entity pointing to itself

+ if you wish to keep "singletons" what would your "current" attrs do?

not itself

why do you think it points to itself, i shared the data from my datascript, none of the currents are pointing to themselves, they are pointing to nested stuff

it is still unclear to me what you are trying to achieve with "current" attrs. when I ask you questions - I understand the answers (I think), but when I get back to your schema and data - those do not match up with my understanding

🙂

well, that's not your fault, i am probably misusing datascript all kinds of ways

ident is just to find it's current, and current is always a nested map with values, the current values for the ident

that was my reasoning

if you have a bunch of singletons, with known "ids/idents", why do you need to link them more?

you don't need this

ident is just a value of attribute

i don't know what you mean by singleton, in my book singleton is a buzzword that just means "object"

key entity - is the id

the objects do not have ids/idents

i have ids/idents and i want to save whatever in them

each entity can have unlimited amount of various attributes

one of which is called :db/ident (people call it this way since datomic has a built in mechanics around it. in datascript it means nothing special)

so

you "history object" is just 1 for example.

yeah

vector("DB_IDENT" "history") is just to find the id

that 1 has an attribute :db/ident = "history". only one entity in db can have a combination of attr=val :db/ident = "history", because you said so in schema: {:db/ident {:db/unique :db.unique/identity}}

originally i started using this because i like that datascript flattens large objects if the schema is done well

so. you can attach the rest of the "destructured" attributes directrly to 1

but then i have to do manual data mangling

it'll look like this:

{:db/id 1
 :db/identity "history"
 :hash "foo"
 :url "bar"
... }

yeah, so actually, it would look like this:

and when you update something signed with db/ident=history - you are updating 1

no need to split it into 2 levels, and link those with "current" (however it can be done)

ds.db_with(db, vector(hashMap(DB_ID, vector("DB_IDENT", "history"), "hash", location.hash, "search", location.search, "pathname", location.pathname)))

and this might be ok with 3 attributes now

but as i said, i liked the fact that datascript can do this work for me

ok, nooow I might see what you mean there

you just need to set :db/ident or :db/id on js object before transacting it

to let datascript know you want to update existing thing, not create new one

yeah, i understood this from your previous explanation 🙂

it's just that i don't know that id, and i don't like changing objects because there comes the evil DEOPTIMIZATION thing in JiT

I just repeated it for myself: first time I wrote it - was a guess answer, but now I feel like it is actual answer 🙂

> it's just that i don't know that id but you said you used idents

on the top level to find the top id, but after that, no. so for history i find 1 or 4 or whatever, but then the nested stuff has 5 or 12 or whatever

and there is no ident for that

mostly because i did not know i need it, and then because i don't like changing objects

i wll now go home because i don't want to be at work in 15 minutes when it will be friday 🙂

thanks for all the great help @misha

i sleep on this and probably delete currents and rewrite stuff

(let [conn (ds/create-conn {:db/ident {:db/unique :db.unique/identity}
                            :history/location {:db/valueType :db.type/ref}})]
  
  (ds/transact! conn [{:db/ident "history"
                       :history/location {:db/ident "location"
                                          :url "foo"
                                          :hash "bar"}}])
  (ds/transact! conn [[:db/add [:db/ident "location"] :hash "baz"]])
  (prn (into [] (ds/datoms @conn :eavt))))

=>
[#datascript/Datom [1 :db/ident "history" 536870913 true]
 <#C07V8N22C|datascript>/Datom [1 :history/location 2 536870913 true]
 <#C07V8N22C|datascript>/Datom [2 :hash "baz" 536870914 true]
 <#C07V8N22C|datascript>/Datom [2 :url "foo" 536870913 true]
 <#C07V8N22C|datascript>/Datom [2 :db/ident "location" 536870913 true]]

if you will have single history and single location, and you want to update location in-place, and keep location linked to history: 1. create 2 idents 2. link them once 3. update just location, providing db/ident to avoid creating new, instead of updating existing one

yeah, this is very close 🙂

and i mean that in a good way