We need to allow overruns during the async multi-broker context spawning
init bc some backends might take longer then others to setup (eg.
binance vs. kucoin) and result in some context (stream) being overrun by
the time we get to the `.open_stream()` phase. Ideally, we can maybe
adjust the concurrent setup to be more of a task-per-provider style to
avoid this in the future - which would also in theory result in
more-immediate per-provider setup in terms showing ready feeds asap.
Also, does a bunch of renaming from fqsn -> fqme and drops the lower
casing of input symbols instead expecting the caller to know what the
data backend it's requesting is going to be able to handle in terms of
symbology.
Instead of pre-converting and mapping piker style fqmes to
`KucoinMktPair`s make `Client._pairs` keyed by the kucoin native market
ids and instead also create a `._fqmes2mktids: bidict[str, str]` for
doing lookups to the native pair from the fqme.
Also, adjust any remaining `fqsn` naming to fqme.
So that styling is preserved on write but requires that we pop `None`
values (in this case any unset `.expiry` transactions) due to `tomkit`
having no support for writing them as values?
So that we aren't creating blank files for legacy configs (as we do name
changes or wtv). Further change `.get_conf_path()` to validate against
new `account.` prefix and a god `conf.toml` file.
Allows for direct loading of an "account file configuration" contents
without having to pass the explicit config dir path. In this case we are
also rewriting the `pps.<brokername>.<accnt_name>.toml` file names to
instead have a `account.` prefix, but providing this helper function
allows such changes more easily in the future - since callers won't have
to use the lower level `.load()` input signature.
Also add some todo comments about moving to `tomlkit`.
I figure we might as well support multiple types of distributed
multi-host setups; why not allow running the API (gateway) and thus vnc
server on a diff host and allowing clients to connect and do their thing
B)
Deatz:
- make `ib._util.data_reset_hack()` take in a `vnc_host` which gets
proxied through to the `asyncvnc` client.
- pull `ib_insync.client.Client` host value and pass-through to data
reset machinery, presuming the vnc server is running in the same
container (and/or the same host).
- if no vnc connection **and** no i3ipc trick can be used, just report
to the user that they need to remove the data throttle manually.
- fix `feed.get_bars()` to handle throttle cases the same based on error
msg matching, not error the code and add a max `_failed_resets` count
to trigger bailing on the query loop.
No longer need to implement connection timeout logic in the streaming
code, instead we just `async for` that bby B)
Further refining:
- better `KucoinTrade` msg parsing and handling with object cases.
- make `subscribe()` do sub request in a loop wand wair for acks.
Previously you couldn't have a brokerd backend which defined
`.trades_dialogue()` but which could also indicate that the paper
clearing engine should be used. This adds that support by allowing the
endpoint task to return a simple `'paper'` string, in which case the ems
will boot a paperboi.
The obvious useful case for this is if you have a broker you want to use
but do not have actual broker credentials setup (yet) with that provider
in your `brokers.toml`; demonstrated here with the adjustment to
`kraken`'s startup to no longer raise a runtime error B)
To fit with the rest of the new requirements added in `.data.validate`
this adds `FeedInit` init including `MktPair` and `Asset` loading for
all spot currencies provided by `kucoin`.
Deatz:
- add a `Currency` struct and accompanying `Client.get_currencies()` for
storing all asset infos.
- implement `.get_mkt_info()` which loads all necessary accounting and
mkt meta-data structs including adding `.price/size_tick` fields to
the `KucoinMktPair`.
- on client boot, async spawn requests to cache both symbols and currencies.
- pass `subscribe()` as the `fixture` arg to `open_autorecon_ws()`
instead of opening it manually.
Other:
- tweak `Client._request` to not expect the prefixed `'/'` for the
`endpoint: str`.
- change the `api_v` arg to just be `api: str`.
Since it's a bit weird having service specific implementation details
inside the general service `._daemon` mod, and since i'd mentioned
trying this re-org; let's do it B)
Requires enabling the new mod in both `pikerd` and `brokerd` and
obviously a bit more runtime-loading of the service modules in the
`brokerd` service eps to avoid import cycles.
Also moved `_setup_persistent_brokerd()` into the new mod since the
naming would place it there even though the implementation really
wouldn't (longer run) since we want to split up `.data.feed` layer
backend-invoked eps into a separate actor eventually from the "actual"
`brokerd` which will be the actor running **only** the trade control eps
(eg. trades_dialogue()` and friends).
`trio`'s internals don't allow for async generator (and thus by
consequence dynamic reset of async exit stacks containing `@acm`s)
interleaving since doing so corrupts the cancel-scope stack. See details
in:
- https://github.com/python-trio/trio/issues/638
- https://trio-util.readthedocs.io/en/latest/#trio_util.trio_async_generator
- `trio._core._run.MISNESTING_ADVICE`
We originally tried to address this using
`@trio_util.trio_async_generator` in backend streaming code but for
whatever reason stopped working recently (at least for me) and it's more
or less implemented the same way as this patch but with more layers and
an extra dep. I also don't want us to have to address this problem again
if/when that lib isn't able to keep up to date with wtv `trio` is
doing..
So instead this is a complete rewrite of the conc design of our
auto-reconnect ws API to move all reset logic and msg relay into a bg
task which is respawned on reset-requiring events: user spec-ed msg recv
latency, network errors, roaming events.
Deatz:
- drop all usage of `AsyncExitStack` and no longer require client code
to (hackily) call `NoBsWs._connect()` on msg latency conditions,
intead this is all done behind the scenes and the user can instead
pass in a `msg_recv_timeout: float`.
- massively simplify impl of `NoBsWs` and move all reset logic into a
new `_reconnect_forever()` task.
- offer use of `reset_after: int` a count value that determines how many
`msg_recv_timeout` events are allowed to occur before reconnecting the
entire ws from scratch again.
Since we have made `MktPair.bs_mktid` mean something else now, change
all the feed setup var names to instead be more representative of the
actual value: `bs_fqme: str` and use the new `MktPair.bs_fqme` where
necessary.
The legacy version was a `dict` of `dicts` vs. now we want to be handed
a `list[FeedInit]`; process both in a factored way.
Drop `FeedInit.bs_mktid` since it's already defined on `.mkt.bs_mktid`
and we don't really need it top level.
More or less a replacement for what @guilledk did with the initial
attempt at a "broker check" type script a while back except in this case
we're going to always run this validation routine and it now uses a new
`FeedInit` struct to ensure backends are delivering the right schema-ed
data during startup. Also allows us to stick deprecation warnings / and
or strict API compat errors all in one spot (at least for live feeds).
Factors out a bunch of `MktPair` related adapter-logic into a new
`.validate.valiate_backend()` which warns to the backend implementer via
log msgs all the problems outstanding. Ideally we do our backend module
endpoint scan-and-complain regarding missing feature support from here
as well (eg. search, broker/trade ctl, ledger processing, etc.).
It needed some work..
- Make `unpack_fqme()` always return a 4-tuple handling the venue and
suffix parts more generally.
- add `Asset.Asset.guess_from_mkt_ep_key()` a like-it-sounds hack at
trying to render a `.dst: Asset` for most most purposes throughout the
stack.
- always try to preprocess the input `fqme: str` with `unpack_fqme()` in
`MktPair.from_fqme()` and use the new `Asset` method (above) to make
up a `.dst: Asset` pulling as much meta-info we can from the caller.
- add `MktPair.bs_fqme` to get the thing without the broker part..
- add an `'unknown'` value to the `_derivs` def.
- drop `Symbol.from_fqsn()` and `unpack_fqsn()` more generally (yes
BREAKING).
- only preload necessary (one for clearing, all for ledger sync)
`MktPair` info from the backend using `.get_mkt_info()`, build the
`mkt_by_fqme: dict[str, MktPair]` and pass it to
`TransactionLedger.iter_trans()`.
- use new `TransactionLedger.update_from_t()` method on clears.
- sanity check all `mkt_by_fqme` entries against `Flume.mkt` values
when we open a data feed.
- rename `PaperBoi._syms` -> `._mkts`.
Turns out we actually had further pp entry bugs due to *not quantizing*
the size inside `.minimize_clears()` method calcs; fix that using
`Position.sys.mkt.quantize()` as is done in `Position.calc_size()`.
Fix `PpTable.write_config()` to drop from the TOML config any
`closed: dict[str, Position]` entries delivered by `.dump_active()`.
Add a more detailed doc string for our position type and a little todo
for the `.bep` B)
Since ledger records are often provided (and thus stored) from most
backends *without* containing the info we normally need for accounting
defined by `MktPair`, this extends the ledger method to take in a table
that allows assigning the `Transaction.sys` from an fqme lookup. This
way client code (like the paper engine and new ledger mgmt tools) can
do the mkt info lookup before hand and then load both ledger
`Transactions` and positions via the `PpTable` and get correct
accounting calculations, always :fingers_crossed:
Also adds `TransactionLedger.update_from_t(t: Transaction)` to allow
updating directly from an existing tran instead of making the user cast
to a `dict` first. Includes fix to `.to_dict()` to always pop the `.sym`
again to avoid client code having to do so.
Wow not sure how that happened, but we should probably use the correct
market precision info for the correct parameter..
Also, use `@lru_cache` on new `get_mkt_info()` ep, seems to work?
In `.feed` and `._sampling` move to using the new
`tractor.Context.open_stream(allow_overruns: bool)` (cough, A BREAKING
CHANGE).
Also set `Flume.mkt` during construction in `.feed.open_feed()`.
Might as well try and flip it over to the new type; make appropriate
dict serialization changes in `.to_msg()`. Alias back to `.symbol:
Symbol` with a property.
Frickin ib, they give you the `0.001` (or wtv) in the
`ContractDetails.minSize: float` but won't accept fractional sizes
through the API.. Either way, it's probably not sane to be supporting
fractional order sizes for legacy instruments by default especially
since it in theory affects a lot of the clearing outcomes by having ib
do wtv magical junk behind the scenes to make it work..
Order mode previously was just willy-nilly sending `float` prices
(particularly on order edits) which are generated from the associated
level line. This actually uses the `MktPair.price_tick: Decimal` to
ensure the value is rounded correctly before submission to the ems..
Also adjusts the order mode init to expect a table of tables of startup
position messages, with the inner table being keyed by fqme per msg.
Tried a couple libs and ended up sticking with `rich` (since it's the
sibling lib to `typer`) but also (initially) implemented a version with
`blessings` that I ended up commenting out (and will likely remove).
Adjusted the CLI I/O a slight bit as well:
- require a fully qualified account name of the form:
`<brokername>.<accountname>` and error on non-matching input.
- dump positions summary lines as humanized size, ppu and cost basis
values per line.
When processing paper trades ledgers we normally won't have specific
`MktPair` info for the backend market we're simulating, as such we
need to look up this info when updating pps.toml files such that we
get precision info correct (particularly in the case of cryptos!) and
can also run paper ledger processing without running the simulated
clearing loop. In order to make it happen we lookup any `get_mkt_info()`
ep on the backend and pass the output to the `force_mkt` input of the
`PpTable.update_from_trans()` method.
This will (likely) act as a new backend query endpoint for other `piker`
(client) code to lookup `MktPair` info from each backend. To start it
also returns the backend-broker's local `Pair` (or wtv other type) as
well.
The main motivation for this is for our paper engine which can require
the mkt info when processing paper-trades ledgers which do not contain
appropriate info to compute position metrics.
Instead of stripping the broker part just use the full fqme for all
`Transaction.bs_mktid: str` values since it makes indexing the `PpTable`
much easier with less key mangling..
Change the root-service-task entrypoint to accept the level and
setup a console log as is now expected for all sub-services. Cast all
backend delivered startup `BrokerdPosition` msgs and log them to
console.
If you want a sub-actor to write console logs (with the right level) the
`get_console_log()` call has to be made somewhere during service task
startup. Previously this wasn't well formalized nor used (depending on
daemon) so passing `loglevel` to the service's root-task-endpoint (eg.
`_setup_persistent_brokerd()`) encourages that the daemon's logging is
configured during init according to the spawner's requesting logging
config. The previous `get_console_log()` call happening inside
`maybe_spawn_daemon()` wasn't actually doing anything in the target
daemon XD, so obviously remove that and instead passthrough loglevel
to the ctx endpoints and service manager methods.
Turns out we don't hookup our eventkit handler until after the
`load_aio_clients()` is complete, which means we can't get
`ib_insync.Client.apiError` events unless inside the asyncio side task.
So I guess try to report any such errors during API scan (note the
duplicate client id case is a special one from ibis itself) even though
we're not going to catch them trio side. The hack to work around this is
to just increment the client id value with the `connect_retries` led `i`
value even though that will break on more then 3 clients attached to an
API endpoint lul ..
Further adjustments that were to the end of trying to fix this proper:
- add `remove_handler_on_err()` cm to disconnect a handler when the trio
side of the channel closes.
- actually connect to client api erros in our `Client.inline_errors()`
- increase connect timeout to a sec.
- change the trio-asyncio proxy response-msg loop over to `match:`
syntax and raise on unhandled msgs from eventkit handlers.
We previously only offered a sync API (which was recently renamed to
`.<meth>_nowait()` style) since initially all order control was from our
`OrderMode` Qt driven UI/UX. This adds the equivalent async methods for
both testing as well as eventual auto-strat driven control B)
Also includes a bunch of renaming:
- `OrderBook` -> `OrderClient`.
- better internal renaming of the client's mem chan vars and add a ref
`._ems_stream: tractor.MsgStream`.
- drop `get_orders()` factory, just always check for the actor-global
instance and always set the ems stream on that client (in case old one
was closed).
This will end up being super handy for testing our accounting subsystems
as well as providing unified and simple cli utils for managing ledgers
and position tracking. Allows loading the paper boi without starting
a data feed and instead just trigger ledger and pps loading without
starting the entire clearing engine.
Deatz:
- only init `PaperBoi` and start clearing loop (tasks) if a non-`None`
fqme is provided, ow just `Context.started()` the existing pps msgs
as loaded from the ledger.
- always update both the ledger and pp table on startup and pass
a single instance of each obj to the `PaperBoi` for reuse (without
opening and closing backing config files since we now have
`.write_config()`).
- drop the global `_positions` dict, it's not needed any more if we use
a `PaperBoi.ppt: PpTable` which persists with the engine actor's
lifetime.
Add a new `class TransactionLedger(collections.UserDict)` for managing
ledger (files) from a `dict`-like API. The main motivations being easy
conversion between `dict` <-> `Transaction` obj forms as well as dynamic
(toml) file updates via a set of methods:
- `.write_config()` to render and write state to the local toml file.
- `.iter_trans()` to allow iterator style conversion to `Transaction`
form for each entry.
- `.to_trans()` for the dict output from the above.
Some adjustments to `Transaction` namely making `.sym/.sys` optional for
now so that paper engine entries can be loaded (offline) without
connecting to the emulated broker backend. Move to using `pathlib.Path`
throughout for bootyful toml file mgmt B)
When loading a `Position` from a pps file we might not have the entire
`MktPair` field-set loaded (though going forward that shouldn't really
ever happen except in the case of a legacy `pps.toml`), in which case we
can check if the `.fqme: str` value loaded from the transaction is
longer and use that instead - presuming it must have more mkt meta-data
filled out.
Also includes some more `fqsn` -> `fqme` renames.
Been meaning to do this port for a while and since it makes passing
around file handles (presumably alongside the in mem obj form) a lot
simpler/nicer and the implementations of all the config file handling
much more terse with less presumptions about the form of filename/dir
`str` values all over the place B)
moar technically, let's us:
- drop remaining `.config` usage of `os.path`.
- return `Path`s from most routines.
- adds a special case to `get_conf_path()` such that if the input name
contains a `pps.` pattern, we avoid validating the name; this is going
to be used by new `.accounting.open_pps()` code which will instead
write a separate TOML file for each account B)
Previous we were re-processing all ledgers for every position msg
received from the API, per client.. Instead do that once in a first pass
and drop all key-miss lookups for `bs_mktid`s; it should never happen.
Better typing for in-routine vars, convert pos msg/objects to `dict`
prior to logging so it's sane to read on console. Skip processing
specifically option contracts for now.
Turns out `binance` is pretty great with their schema since they have
more or less the same data schema for their exchange info ep which we
wrap in a `Pair` struct:
https://binance-docs.github.io/apidocs/spot/en/#exchange-information
That makes it super easy to provide the most general case for filling
out a `MktPair` with both `.src/dst: Asset` to maintain maximum
meta-data B)
Deatz:
- adjust `Pair` to have `.size/price_tick: Decimal` by parsing out
the values from the filters field; TODO: we should probably just rewrite
the input `.filter` at init time so we can keep the frozen style.
- rename `Client.mkt_info()` (was `.symbol_info` to `.exch_info()`
better matching the ep name and have it build, cache, and return
a `dict[str, Pair]`; allows dropping `.cache_symbols()`
- only pass the `mkt_info: MktPair` field in the init msg!
Accept a msg with any of:
- `.src: Asset` and `.dst: Asset`
- `.src: str` and `.dst: str`
- `.src: Asset` and `.dst: str`
but not the final combo tho XD
Also, fix `.key` to properly cast any `.src: Asset` to string!
If user has loaded from a flex report then we don't want the API records
from the same period to override those; instead just update with any
missing fields from the API schema.
Also, always `str`-ify the contract id (what is set for the `.bs_mktid`
*before* packing into transaction type to ensure when serialized to
`pps.toml` there are no discrepancies at the codec level.. smh
Instead adjust `load_aio_clients()` to only reload clients detected as
non-loaded or disconnected (2 birds), and avoid use of the global module
table which could result in stale disconnected clients persisting on
multiple `brokerd` client reconnects, resulting in error.
To make nested `msgspec.Struct`s work we need to tell the codec that the
`.symbol` is some struct def, since we don't really need to enforce that
(yet) we're just going to enc/dec as `str` until we further formalize
and/or need something more complex.
Initial attempt at getting the sampling and shm layer to use the new mkt
info meta-data type. Draft out a potential `BackendInitMsg:
msgspec.Struct` for validating the init msg returned from the
`stream_quotes()` start value; obvs don't actually use it yet.
To be compat with the `Symbol` (for now) and generally allow for reading
the (derivative) contract specific part of the fqme. Adjust
`contract_info: list[str]` and make `src: str = ''` by default.
Add `MktPair` handling block for when a backend delivers
a `mkt_info`-field containing init msg. Adjust the original
`Symbol`-style `'symbol_info'` msg processing to do `Decimal` defaults
and convert to `MktPair` including slapping in a hacky `_atype: str`
field XD
General initial name changes to `bs_mktid` and `_fqme` throughout!
For `price_tick` and `size_tick` we read in `str` and decimal-ize
and now correctly fail over to default values of the same type..
Also, always treat `bs_mktid` field as a `str` in TOML form.
Drop the strange `clears: dict` var from the loading code (not sure why
that was left in smh) and better name `toml_clears_list` for the
TOML-loaded-pre-transaction sequence.
Handle case where `'dst'` field is just a `str` (in which case delegate to
`.from_fqme()`) as well as do `Asset` loading and use our
`Struct.copy()` to enforce type-casting to (for eg. `Decimal`s) such
that we'll now capture typing errors despite IPC transport.
Change `Symbol.tick_size` and `.lot_tick_size` defaults to decimal
for proper casting and type `MktPair.atype: str` since `msgspec` can't
cast to `AssetTypeName` without special handling..
Allows building a `MktPair` from the backend specific `Pair` for
eventual use in the data feed layer. Also adds `Pair.price/tick_size` to
get to the expected tick precision info format.
Add a logic branch for now that switches on an instance check.
Generally swap over all `Position.symbol` and `Transaction.sym` refs to
`MktPair`. Do a wholesale rename of all `.bsuid` var names to
`.bs_mktid`.
Instead let's name it `.sys` for "system", the thing we use to conduct
the "transactions" ..
Also rename `.bsuid` -> `.bs_mktid` for "backend system market id`
which is more explicit, easier to remember and read.
Prepping to entirely replace `Symbol`; this adds a buncha docs/comments,
better implementation for representing and parsing the FQME: "fully
qualified market endpoint".
Deatz:
- make `.src` an optional field until we figure out how we're going
to support loading source assets from all backends sensibly..
- implement `MktPair.fqme: str` (what was previously called `fqsn`)
using a new util func: `maybe_cons_tokens()`.
- `Symbol.brokers` and expect only `.broker` usage.
- remap anything with `fqsn` in the name to `fqme` with aliases from the
old name.
- implement `unpack_fqme()` with `match:` syntax B)
- add `MktPair.tick_size_digits`, `.lot_size_digits`, `.fqsn`, `.key` for
backward compat.
- make all fqme generation related fields empty `str`s by default.
- add `MktPair.resolved: bool` a flag indicating whether or not `.dst`
is an `Asset` instance or just a string and, `.bs_mktid` the field
to hold the "backend system market id" per broker.
Try out using our new internal type for storing info about kraken's asset
infos now stored in the `Client.assets: dict[str, Asset]` table. Handle
a server error when requesting such info msgs.
Drop everything we can in terms of methods and attrs from `Symbol`:
- kill `.tokens()`, `.front_feed()`, `.tokens()`, `.nearest_tick()`,
`.front_fqsn()`, instead moving logic from these methods into
dependents (and obviously removing any usage from rest of code base,
coming in follow up commits).
- rename `.quantize_size()` -> `.quantize()`.
- re-implement `.brokers`, `.lot_size_digits`, `.tick_size_digits` as
`@property` methods; for the latter two, allows us to minimize to only
accepting min tick decimal values on alternative constructor class
methods and to drop the equivalent instance vars.
- map `_fqsn` related variable names to new and preferred `_fqme`.
We also juggle around some utility functions, moving limited precision
related `decimal.Decimal` routines to the top of module and soon-to-be
legacy `fqsn` related routines to the bottom.
`MktPair` draft type extensions:
- drop requirements for `src_type`, and offer the optional `.dst_type`
field as either a `str` or (new `typing.Literal`) `AssetTypeName`.
- define an equivalent `.quantize()` as (re)defined in `Symbol` but with
`quantity_type: str` field which specifies whether to use the price or
the size precision.
- add a lot more docs, a `.key` property for the "symbol" name, draft
property for a `.fqme: str`
- allow `.src` and `.dst` to be of type `str | Asset`
Add a new `Asset` to capture "things which can be used in markets and/or
transactions" XD
- defines a `.name`, `.atype: AssetTypeName` a financial category tag, `tx_tick:
Decimal` the precision limit for transactions and of course
a `.quantime()` method for doing accounting arithmetic on a given tech
stack.
- define the `atype: AssetTypeName` type as a finite set of `str`s
expected to be used in various ways for default settings in other
parts of the data and order control layers..
Our issue was not having the correct value set on each
`Symbol.lot_tick_size`.. and then doing PPU calcs with the default set
for legacy mkts..
Also,
- actually write `pps.toml` on broker mode exit.
- drop `get_likely_pair()` and import from pp module.
Not sure how this worked before but, the PPU calculation critically
requires that the order of clearing transactions are in the correct
chronological order! Fix this by sorting `trans: dict[str, Transaction]`
in the `PpTable.update_from_trans()` method.
Also, move the `get_likely_pair()` parser from the `kraken` backend here
for future use particularly when we revamp the asset-transaction
processing layer.
Apparently it will likely fix our `trio`-cancel-scopes-corrupted crash
when we try to let our `._web_bs.NoBsWs` do reconnect logic around
the asyn-generator implemented data-feed streaming routines in `binance`
and `kraken`. See the project docs for deatz; obvs we add the lib as
a dep.
Solve this by always scaling the y-range for the major/target curve
*before* the final overlay scaling loop; this implicitly always solve
the case where the major series is the only one in view.
Tidy up debug print formatting and add some loop-end demarcation comment
lines.
This is particularly more "good looking" when we boot with a pair that
doesn't have historical 1s OHLC and thus the fast chart is empty from
outset. In this case it's a lot nicer to be already zoomed to
a comfortable preset number of "datums in view" even when the history
isn't yet filled in.
Adjusts the chart display `Viz.default_view()` startup to explicitly
ensure this happens via the `do_min_bars=True` flag B)
Not sure how i missed this (and left in handling of `list.remove()` and
it ever worked for that?) after the `samplerd` impl in 5ec1a72 but, this
adjusts the remove-broken-subscriber loop to catch the correct
`set.remove()` exception type on a missing (likely already removed)
subscription entry.
For the purposes of eventually trying to resolve last-step indexing
synchronization (an intermittent but still existing) issue(s) that can
happen due to races during history frame query and shm writing during
startup. In fact, here we drop all `hist_viz` info queries from the main
display loop for now anticipating that this code will either be removed
or improved later.
Again, as per the signature change, never expect implicit time step
calcs from overlay processing/machinery code. Also, extend the debug
printing (yet again) to include better details around
"rescale-due-to-minor-range-out-of-view" cases and a detailed msg for
the transform/scaling calculation (inputs/outputs), particularly for the
cases when one of the curves has a lesser support.
As per the change to `slice_from_time()` this ensures this `Viz` always
passes its self-calculated time indexing step size to the time slicing
routine(s).
Further this contains a slight impl tweak to `.scalars_from_index()` to
slice the actual view range from `xref` to `Viz.ViewState.xrange[1]` and
then reading the corresponding `yref` from the first entry in that
array; this should be no slower in theory and makes way for further
caching of x-read-range to `ViewState` opportunities later.
There's been way too many issues when trying to calculate this
dynamically from the input array, so just expect the caller to know what
it's doing and don't bother with ever hitting the error case of
calculating and incorrect value internally.
When the target pinning curve (by default, the dispersion major) is
shorter then the pinned curve, we need to make sure we find still find
the x-intersect for computing returns scalars! Use `Viz.i_from_t()` to
accomplish this as well and, augment that method with a `return_y: bool`
to allow the caller to also retrieve the equivalent y-value at the
requested input time `t: float` for convenience.
Also tweak a few more internals around the 'loglin_ref_to_curve'
method:
- only solve / adjust for the above case when the major's xref is
detected as being "earlier" in time the current minor's.
- pop the major viz entry from the overlay table ahead of time to avoid
a needless iteration and simplify the transform calc phase loop to
avoid handling that needless cycle B)
- add much better "organized" debug printing with more clear headers
around which "phase"/loop the message pertains and well as more
explicit details in terms of x and y-range values on each cycle of
each loop.
Previously when very zoomed out and using the `'r'` hotkey the
interaction handler loop wouldn't trigger a re-(up)sampling to get
a more detailed curve graphic and instead the previous downsampled
(under-detailed) graphic would show. Fix that by ensuring we yield back
to the Qt event loop and do at least a couple render cycles with paired
`.interact_graphics_cycle()` calls.
Further this flips the `.start/signal_ic()` methods to use the new
`.reset_graphics_caches()` ctr-mngr method.
Instead delegate directly to `Viz.default_view()` throughout charting
startup and interaction handlers.
Also add a `ChartPlotWidget.reset_graphics_caches()` context mngr which
resets all managed graphics object's cacheing modes on enter and
restores them on exit for simplified use in interaction handling code.
This finally seems to mitigate all the "smearing" and "jitter" artifacts
when using Qt's "coordinate cache" graphics-mode:
- whenever we're in a mouse interaction (as per calls to
`ChartView.start/signal_ic()`) we simply disable the caching mode (set
`.NoCache` until the interaction is complete.
- only do this (for now) during a pan since it doesn't seem to be an
issue when zooming?
- ensure disabling all `Viz.graphics` and `.ds_graphics` to be agnostic
to any case where there's both a zoom and a pan simultaneously (not
that it's easy to do manually XD) as well as solving the problem
whenever an OHLC series is in traced-and-downsampled mode (during low
zoom).
Impl deatz:
- rename `ChartView._ic` -> `._in_interact: trio.Event`
- add `.ChartView._interact_stack: ExitStack` which we use to open.
and close the `FlowGraphics.reset_cache()` mngrs from mouse handlers.
- drop all the commented per-subtype overrides for `.cache_mode: int`.
- write up much better doc strings for `FlattenedOHLC` and `StepCurve`
including some very basic ASCII-art diagrams.
When the minor has the same scaling as the major in a given direction we
should still do back-scaling against the major-target and previous
minors to avoid strange edge cases where only the target-major might not
be shifted correctly to show an matched intersect point? More or less
this just meant making the y-mxmn checks interval-inclusive with
`>=`/`<=` operators.
Also adds a shite ton of detailed comments throughout the pin-to-target
method blocks and moves the final major y-range call outside the final
`scaled: dict` loop.
For the "pin to target major/target curve" overlay method, this finally
solves the longstanding issue of ensuring that any new minor curve,
which requires and increase in the major/target curve y-range, also
re-scales all previously scaled minor curves retroactively. Thus we now
guarantee that all minor curves are correctly "pinned" to their
target/major on their earliest available datum **and** are all kept in
view.
Yah yah, i know it's the same as before (the N > 2 curves case with
out-of-range-minor rescaling the previously scaled curves isn't fixed
yet...) but, this is a much better and optional implementation in less
code. Further we're now better leveraging various new cached properties
and methods on `Viz`.
We now handle different `overlay_technique: str` options using `match:`
syntax in the 2ndary scaling loop, stash the returns scalars per curve
in `overlay_table`, and store and iterate the curves by dispersion
measure sort order.
Further wrt "pin-to-target-curve" mode, which currently still pins to
the largest measured dispersion curve in the overlay set:
- pop major Ci overlay table entries at start for sub-calcs usage when
handling the "minor requires major rescale after pin" case.
- (finally) correctly rescale the major curve y-mxmn to whatever the
latest minor overlay curve by calcing the inverse transform from the
minor *at that point*:
- the intersect point being that which the minor has starts support on
the major's x-domain* using the new `Viz.scalars_from_index()` and,
- checking that the minor is not out of range (versus what the major's
transform calcs it to be, in which case,
- calc the inverse transform from the current out-of-range minor and
use it to project the new y-mxmn for the major/target based on the
same intersect-reference point in the x-domain used by the minor.
- always handle the target-major Ci specially by only setting the
`mx_ymn` / `mx_ymn` value when iterating that entry in the overlay
table.
- add todos around also doing the last sub-sub bullet for all previously
major-transform scaled minor overlays (this is coming next..i hope).
- add a final 3rd overlay loop which goes through a final `scaled: dict`
to apply all range values to each view; this is where we will
eventually solve that last edge case of an out-of-range minor's
scaling needing to be used to rescale already scaled minors XD
In an effort to make overlay calcs cleaner and leverage caching of view
range -> dispersion measures, this adds the following new methods:
- `._dispersion()` an lru cached returns scalar calculator given input
y-range and y-ref values.
- `.disp_from_range()` which calls the above method and returns variable
output depending on requested calc `method: str`.
- `.i_from_t()` a currently unused cached method for slicing the
in-view's array index from time stamp (though not working yet due to
needing to parameterize the cache by the input `.vs.xrange`).
Further refinements/adjustments:
- rename `.view_state: ViewState` -> `.vs`.
- drop the `.bars_range()` method as it's no longer used anywhere else
in the code base.
- always set the `ViewState.in_view: np.ndarray` inside `.read()`.
- return the start array index (from slice) and `yref` value @ `xref`
from `.scalars_from_index()` to aid with "pin to curve" rescaling
caused by out-of-range pinned-minor curves.
Not sure why this was ever allowed but, for slicing to the sample
*before* whatever target time stamp is passed in we should definitely
not return the prior index as for the slice start since that might
include a very large gap prior to whatever sample is scanned to have
the earliest matching time stamp.
This was essential to fixing overlay intersect points searching in our
``ui.view_mode`` machinery..
Adds a small struct which is used to track the most recently viewed
data's x/y ranges as well as the last `Viz.read()` "in view" array data
for fast access by chart related graphics processing code, namely view
mode overlay handling.
Also adds new `Viz` interfaces:
- `Viz.ds_yrange: tuple[float, float]' which replaces the previous
`.yrange` (now set by `.datums_range()` on manual y-range calcs) so
that the m4 downsampler can set this field specifically and then it
get used (when available) by `Viz.maxmin()`.
- `Viz.scalars_from_index()` a new returns-scalar generator which can be
used to calc the up and down returns values (used for scaling overlay
y-ranges) from an input `xref` x-domain index which maps to some
`Ci(xref) = yref`.
It was getting waayy to long to be jammed in a method XD
This moves all the chart-viz iteration and transform logic into a new
`piker.ui.view_mode.overlay_viewlists()` core routine which will make it
a lot nicer for,
- AOT compilation via `numba` / `cython` / `mypyc`.
- decoupling from the `pyqtgraph.ViewBox` APIs if we ever decide to get
crazy and go without another graphics engine.
- keeping your head clear when trying to rework the code B)
As part of solving a final bullet-issue in #455, which is specifically
a case:
- with N > 2 curves, one of which is the "major" dispersion curve" and
the others are "minors",
- we can run into a scenario where some minor curve which gets pinned to
the major (due to the original "pinning technique" -> "align to
major") at some `P(t)` which is *not* the major's minimum / maximum
due to the minor having a smaller/shorter support and thus,
- requires that in order to show then max/min on the minor curve we have
to expand the range of the major curve as well but,
- that also means any previously scaled (to the major) minor curves need
to be adjusted as well or they'll not be pinned to the major the same
way!
I originally was trying to avoid doing the recursive iteration back
through all previously scaled minor curves and instead decided to try
implementing the "per side" curve dispersion detection (as was
originally attempted when first starting this work). The idea is to
decide which curve's up or down "swing in % returns" would determine the
global y-range *on that side*. Turns out I stumbled on the "align to
first" technique in the process: "for each overlay curve we align its
earliest sample (in time) to the same level of the earliest such sample
for whatever is deemed the major (directionally disperse) curve in
view".
I decided (with help) that this "pin to first" approach/style is equally
as useful and maybe often more so when wanting to view support-disjoint
time series:
- instead of compressing the y-range on "longer series which have lesser
sigma" to make whatever "shorter but larger-sigma series" pin to it at
an intersect time step, this instead will expand the price ranges
based on the earliest time step in each series.
- the output global-returns-overlay-range for any N-set of series is equal to
the same in the previous "pin to intersect time" technique.
- the only time this technique seems less useful is for overlaying
market feeds which have the same destination asset but different
source assets (eg. btceur and btcusd on the same chart since if one
of the series is shorter it will always be aligned to the earliest
datum on the longer instead of more naturally to the intersect sample
level as was in the previous approach).
As such I'm going to keep this technique as discovered and will later
add back optional support for the "align to intersect" approach from
previous (which will again require detecting the highest dispersion
curve direction-agnostic) and pin all minors to the price level at which
they start on the major.
Further details of the implementation rework in
`.interact_graphics_cycle()` include:
- add `intersect_from_longer()` to detect and deliver a common datum
from 2 series which are different in length: the first time-index
sample in the longer.
- Rewrite the drafted `OverlayT` to only compute (inversed log-returns)
transforms for a single direction and use 2 instances, one for each
direction inside the `Viz`-overlay iteration loop.
- do all dispersion-per-side major curve detection in the first pass of
all `Viz`s on a plot, instead updating the `OverlayT` instances for
each side and compensating for any length mismatch and
rescale-to-minor cases in each loop cycle.
Previously we were aligning the child's `PlotItem` to the "root" (top
most) overlays `ViewBox`..smh. This is why there was a weird gap on the
LHS next to the 'left' price axes: something weird in the implied axes
offsets was getting jammed in that rect.
Also comments out "the-skipping-of" moving axes from the overlay's
`PlotItem.layout` to the root's linear layout(s) when an overlay's axis
is read as not visible; this isn't really necessary nor useful and if we
want to remove the axes entirely we should do it explicitly and/or
provide a way through the `ComposeGridLayout` API.
Despite there being artifacts when interacting, the speedups when
cross-hair-ing are just too good to ignore. We can always play with
disabling caches when interaction takes place much like we do with feed
pausing.
When zoomed in alot, and thus a quote driven y-range resize takes place,
it makes more sense to increase the `range_margin: float` input to
`._set_yrange()` to ensure all L1 labels stay in view; generally the
more zoomed in,
- the smaller the y-range is and thus the larger the needed margin (on
that range's dispersion diff) would be,
- it's more likely to get a last datum move outside the previous range.
Also, always do overlayT style scaling on the slow chart whenever it
treads.
Since it can be desirable to dynamically adjust inputs to the y-ranging
method (such as in the display loop when a chart is very zoomed in), this
adds such support through a new `yrange_kwargs: dict[Viz, dict]` which
replaces the `yrange` tuple we were passing through prior. Also, adjusts
the y-range margin back to the original 0.09 of the diff now that we can
support dynamic control.
If there is a common `PlotItem` used for a set of `Viz`/curves (on
a given view) we don't need to do overlay scaling and thus can also
short circuit the viz iteration loop early.
Somewhat of a facepalm but, for incremental update of the auto-yrange
from quotes in the display loop obviously we only want to update the
associated `Viz`/viewbox for *that* fqsn. Further we don't need to worry
about the whole "tick margin" stuff since `._set_yrange()` already adds
margin to the yrange by default; thus we remove all of that.
When the caller passes `do_overlay_scaling=False` we skip the given
chart's `Viz` iteration loop, and set the yrange immediately, then
continue to the next chart (if `do_linked_charts` is set) instead of
a `continue` short circuit within the viz sub-loop.
Deats:
- add a `_maybe_calc_yrange()` helper which makes the `yranges`
provided-or-not case logic more terse (factored).
- add a `do_linked_charts=False` short circuit.
- drop the legacy commented swing % calcs stuff.
- use the `ChartView._viz` when `do_overlay_scaling=False` thus
presuming that we want to handle the viz mapped to *this* view box.
- add a `._yrange` "last set yrange" tracking var which keeps record of
the last ymn/ymx value set in `._set_yrange()` BEFORE doing range
margins; this will be used for incremental update in the display loop.
Since each symbol's feed is multiplexed by quote key (an fqsn), we can
avoid scaling overlay curves on any single update, presuming each quote
driven cycle will trigger **only** the specific symbol's curve.
Also disables fsp `.interact_graphics_cycle()` calls for now since it
seems they aren't really that critical to and we should be using the
same technique as above (doing incremental y-range checks/updates) for
FSPs as well.
The reason (fsp) subcharts were not linked-updating correctly was
because of the early termination of the interact update loop when only
one "overlay" (aka no other overlays then the main curve) is detected.
Obviously in this case we still need to iterate all linked charts in the
set (presuming the user doesn't disable this).
Also tweaks a few internals:
- rename `start_datums: dict` -> `overlay_table`.
- compact all "single curve" checks to one logic block.
- don't collect curve info into the `overlay_table: dict` when
`do_overlay_scaling=True`.
Such that we still y-range auto-sort inside
`ChartView.interact_graphics_cycle()` still runs on the unit vlm axis
and we always size such that the y-label stays in view.
Since we pretty much always want the 'bottom' and any side that is not
declared by the caller move the axis hides into this method. Lets us
drop the same calls in `.ui._fsp` and `._display`.
This also disables the auto-ranging back-linking for now since it
doesn't seem to be working quite yet?
In situations where clients are (dynamically) subscribing *while*
broadcasts are starting to taking place we need to handle the
`set`-modified-during-iteration case. This scenario seems to be more
common during races on concurrent startup of multiple symbols. The
solution here is to use another set to take note of subscribers which
are successfully sent-to and then skipping them on re-try.
This also contains an attempt to exception-handle throttled stream
overruns caused by higher frequency feeds (like binance) pushing more
quotes then can be handled during (UI) client startup.
This was a subtle logic error when building the `plots: dict` we weren't
adding the "main (ohlc or other source) chart" from the `LinkedSplits`
set when interacting with some sub-chart from `.subplots`..
Further this tries out bypassing `numpy.median()` altogether by just
using `median = (ymx - ymn) / 2` which should be nearly the same?
In the (incrementally updated) display loop we have range logic that is
incrementally updated in real-time by streams, as such we don't really
need to update all linked chart's (for any given, currently updated
chart) y-ranges on calls of each separate (sub-)chart's
`ChartView.interact_graphics_cycle()`. In practise there are plenty of
cases where resizing in one chart (say the vlm fsps sub-plot) requires
a y-range re-calc but not in the OHLC price chart. Therefore
we always avoid doing more resizing then necessary despite it resulting
in potentially more method call overhead (which will later be justified
by better leveraging incrementally updated `Viz.maxmin()` and
`media_from_range()` calcs).
A super snappy `numpy.median()` calculator (per input range) which we
slap an `lru_cache` on thanks to handy dunder method hacks for such
things on mutable types XD
use the new `do_overlay_scaling: bool` since we know each feed will have
its own updates (cuz multiplexed by feed..) and we can avoid
ranging/scaling overlays that will make their own calls.
Also, pass in the last datum "brighter" color for ohlc curves as it was
originally (and now that we can pass that styling bit through).
Facepalm, obviously absolute array indexes are not going to necessarily
align vs. time over multiple feeds/history. Instead use
`np.searchsorted()` on whatever curve has the smallest support and find
the appropriate index of intersection in time so that alignment always
starts at a sensible reference.
Also adds a `debug_print: bool` input arg which can enable all the
prints when working on this.
We can determine the major curve (in view) in the first pass of all
`Viz`s so drop the 2nd loop and thus the `mxmn_groups: dict`. Also
simplifies logic for the case of only one (the major) curve in view.
Turns out this is a limitation of the `ViewBox.setYRange()` api: you
can't call it more then once and expect anything but the first call to
be applied without letting a render cycle run. As such, we wait until
the end of the log-linear scaling loop to finally apply the major curves
y-mx/mn after all minor curves have been evaluated.
This also drops all the debug prints (for now) to get a feel for latency
in production mode.
We ended up doing groups maxmin sorting at the interaction layer (new
the view box) and thus this method is no longer needed, though it was
the reference for the code now in `ChartView.interact_graphics_cycle()`.
Further this adds a `remove_axes: bool` arg to `.insert_plotitem()`
which can be used to drop axis entries from the inserted pi (though it
doesn't seem like we really ever need that?) and does the removal in
a separate loop to avoid removing axes before they are registered in
`ComposedGridLayout._pi2axes`.
When there are `N`-curves we need to consider the smallest
x-data-support subset when figuring out for each major-minor pair such
that the "shorter" series is always returns aligned to the longer one.
This makes the var naming more explicit with `major/minor_i_start` as
well as clarifies more stringently a bunch of other variables and
explicitly uses the `minor_y_intersect` y value in the scaling transform
calcs. Also fixes some debug prints.
In very close manner to the original (gut instinct) attempt, this
properly (y-axis-vertically) aligns and scales overlaid curves according
to what we are calling a "log-linearized y-range multi-plot" B)
The basic idea is that a simple returns measure (eg. `R = (p1 - p0)
/ p0`) applied to all curves gives a constant output `R` no matter the
price co-domain in use and thus gives a constant returns over all assets
in view styled scaling; a intuitive visual of returns correlation. The
reference point is for now the left-most point in view (or highest
common index available to all curves), though we can make this
a parameter based on user needs.
A slew of debug `print()`s are left in for now until we iron out the
remaining edge cases to do with re-scaling a major (dispersion) curve
based on a minor now requiring a larger log-linear y-range from that
previous major' range.
In the dispersion swing calcs, use the series median from the in-view
data to determine swing proportions to apply on each "minor curve"
(series with lesser dispersion the one with the greatest). Track the
major `Viz` as before by max dispersion. Apply the dispersion swing
proportions to each minor curve-series in a third loop/pass of all
overlay groups: this ensures all overlays are dispersion normalized in
their ranges but, minor curves are currently (vertically) centered (vs.
the major) via their medians.
There is a ton of commented code from attempts to try and vertically
align minor curves to the major via the "first datum" in-view/available.
This still needs work and we may want to offer it as optional.
Also adds logic to allow skipping margin adjustments in `._set_yrange()`
if you pass `range_margin=None`.
On overlaid ohlc vizs we compute the largest max/min spread and
apply that maxmimum "up and down swing" proportion to each `Viz`'s
viewbox in the group.
We obviously still need to clip to the shortest x-range so that
it doesn't look exactly the same as before XD
We were hacking this before using the whole `ChartView._maxmin()`
setting stuff since in some cases you might want similarly ranged paths
on the same view, but of course you need to max/min them together..
This adds that group sorting by using a table of `dict[PlotItem,
tuple[float, float]` and taking the abs highest/lowest value for each
plot in the viz interaction update loop.
Also removes the now commented signal registry calls and thus
`._yranger`, drops the `set_range: bool` from `._set_yrange` and adds
and extra `.maybe_downsample_graphics()` to the mouse wheel handler to
avoid a weird slow debounce where ds-ing is delayed until a further
interaction.
It's kind of hard to understand with the C++ fan-out to multiple views
(imo a cluster-f#$*&) and seems honestly just plain faster to loop (in
python) through all the linked view handlers XD
Core adjustments:
- make the panning and wheel-scroll handlers just call
`.maybe_downsample_graphics()` directly; drop all signal emissions.
- make `.maybe_downsample_graphics()` loop through all vizs per subchart
and use the new pipeline-style call sequence of:
- `Viz.update_graphics() -> <read_slc>: tuple`
- `Viz.maxmin(i_read_range=<read_slc>) -> yrange: tuple`
- `Viz.plot.vb._set_yrange(yrange=yrange)`
which inlines all the necessary calls in the most efficient way whilst
leveraging `.maxmin()` caching and ymxmn-from-m4-during-render to
boot.
- drop registering `._set_yrange()` for handling `.sigRangeChangedManually`.
Computes the maxmin values for each underlying plot's in-view range as
well as the max up/down swing (in percentage terms) from the plot with
most dispersion and returns a all these values plus a `dict` of plots to
their ranges as part of output.
This broke non-disti-mode actor tree spawn / runtime, seemingly because
the cli entrypoint for a `piker chart` also sends these values down
through the call stack independently? Pretty sure we don't need to send
the `enable_modules` from the chart actor anyway.
Needed to move the startup sequence inside the `try:` block to guarantee
we always do the (now shielded) `.cancel()` call if we get a cancel
during startup.
Also, support an optional `started_afunc` field in the config if
backends want to just provide a one-off blocking async func to sync
container startup. Add a `drop_root_perms: bool` to allow persisting
sudo perms for testing or dyanmic container spawning purposes.
Provides a more correct solution (particularly for distributed testing)
to override the `piker` configuration directory by reading the path from
a specific `tractor._state._runtime_vars` entry that can be provided by
the test harness.
Also fix some typing and comments.
Tyler Goodlet
jaredgoldman