Repository navigation
v0.3.0
The release that stops the Air strip asserting its own answer. Twenty-one commits since v0.2.0: a second air model that computes the flow instead of being told it, a title block that can be signed, a station reader that refuses what it cannot read rather than fatalling the run, and a heat balance that says out loud which way it points. No breaking changes; every v0.2.0 permalink still decodes.
Features
-
An AirflowNetwork air model beside the scheduled one (#43): the Air strip now states its subject two ways. Either you state an air change rate, or a pressure network computes the flow from wind and stack effect at every timestep, out of an envelope leakiness and an openable area per wall — which is what lets the desk answer whether it matters that the openable window is on the west elevation rather than the south. The engine simulates one or the other and never both, so the model that is out has its objects deleted rather than zeroed. Costs about 20 ms on a design day and 3.1x on a year.
-
The title block can be signed (#40): type a name into
Drawn byand every model the desk hands over carries it, at the head of the IDF and in the run bundle'sMANIFEST.txt. The cell is empty until you fill it — a title block names whoever drew that drawing, and here that is whoever is at the console. The name is kept in this browser and never in a link, so a scheme you send carries the building and not whoever drew it. Signing runs nothing and moves no reading. -
The page says who built it: the head's stamp gains a Toolkit row beside Engine, Runtime and Upload, so the page states what it is a demo of without anyone opening the Disclosure band.
-
The scoreboard's em dashes can be pressed: a desk that ships free-running used to meet nine em dashes and the same grey sentence beside each one. One note above the table now carries the fix as a press, escalating from "patch System in" to "attach a weather file", and the kept-schemes shelf carries the same offer.
-
Page views are counted through GoatCounter — no cookie, no identifier, nothing cross-site, and the URL fragment is not recorded, so a shared desk never lands in the table as its own page.
Fixes
-
A station whose design conditions cannot be read is refused, instead of fatalling the run (#42). Searching
Bostonput WMO 994971 at the top of the list, and attaching it stopped the engine before a single hour was simulated: the site publishes no annual cooling condition at all, so the reader fell back to the first day of the right season — a January day — carrying the literal textNwhere onebuilding had no number to publish. EnergyPlus rejectsNin a numeric field, and the first thing a new reader did on this page could be watch it fail.The obvious fix was the wrong one: the name the reader used to ask for,
Ann Clg 1% Condns DB=>MWB, is published for only 69 sites in 120, so simply requiring it would have refused 51 of them, 39 working perfectly. The sheet now declares an ordered list of annual design days it will accept, at the severity it has always claimed, and takes the first one published whose numbers parse. Measured over the same sample: 108 attach, 12 are refused, and every one of the 12 is a marine99xxxxstation publishing no annual cooling condition whatsoever. -
The plate letters the design day it actually got. Any non-winter day was lettered
1% clg db, so 39 sites in 120 were being sized on a.4%dewpoint day under a label claiming otherwise. -
A refused station says what to do instead, reopening the picker on the nearest other stations — two clicks from the refusal to a solved Boston-Logan.
-
A link naming an unusable station says why. The generic "the whole link was set aside" was being lettered over the specific sentence the attach had just written into the same status line.
-
The revisions section no longer overflows a narrow screen. A wide unwrapped code block was winning the grid's default
min-width: autoand pushing the whole block past the viewport instead of scrolling inside its ownoverflow.
Changes to how the sheet reads
-
The heat balance says which way it points (#38). Direction was carried only by the hue and by the absence of a minus sign, which is a colour-only encoding of the one fact the rail exists to state: in monochrome, under forced colours, or read aloud as a swatch, a positive term said nothing whatever about being positive. The convention is now stated under the rail's head, and every rail figure carries
inoroutbeside its watts. The head also says what the±is — one side of the balance, not the two netted. -
Two rail readings say on the strip that they are not the variable they are named after. System air transfer is reported per building and divided back down. Air energy storage enters the balance negated and is relettered Air energy release, because under the variable's own name a reading of −400 W said "storage" and "out" in one breath. Neither transformation was wrong; neither was checkable.
-
A control belonging to the model that is out is no longer drawn (#44). Dimming a withdrawn block said "inert right now" where the truth is "belongs to the other instrument", and a dimmed slider is still draggable, still in the tab order and still moves under an arrow key. Applied to all three channels with that shape: Air's two models, Glazing's simple unit against its layered one, and the rooflights' own glass. A control whose reviving control is adjacent is still dimmed rather than withdrawn.
-
The run log stays in view down the sheet, so a solve fired from the scoreboard no longer reports its stages and wall clock to a margin scrolled off the top.
-
The lede leads with the capability rather than with Denver, since the first climate named is the one a reader takes the tool to be about.
-
The title block says there is no simulation server, rather than
Upload: None— which on a page that simulates a building read as an empty field, and so as an affordance this page does not have. -
The scoreboard and the kept-schemes shelf read as tables again: each standard's name heads a banded section, and each shelf figure is headed over its own column instead of over the empty delta beside it.
Deployment
- Only a tag publishes the site (#50).
shoebox.idfkit.comis a released issue of the drawing and nothing else, for the same reason the title block carries a revision at all. Every other push tomainnow publishes the development channel atshoebox.idfkit.com/dev/, and opening a pull request publishes a preview atshoebox.idfkit.com/<number>/. A release is the same artefact that has been served under/dev/since it was merged, moved to the root.
Upgrade notes
Nothing to do. This is a web page, not a package: open shoebox.idfkit.com and you have it. Permalinks minted under v0.2.0 still decode — the link format is still v1, since the new controls are additions under delta encoding and every new channel ships bypassed.