atsfill 0.0.6.dev203__tar.gz → 0.0.6.dev205__tar.gz

{atsfill-0.0.6.dev203 → atsfill-0.0.6.dev205}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: atsfill
-Version: 0.0.6.dev203
+Version: 0.0.6.dev205
 Summary: Read, dry-run and submit job application forms on hiring systems: Ashby, Greenhouse, Teamtailor, Welcome to the Jungle.
 Keywords: job-applications,ats,ashby,greenhouse,teamtailor,ai-agents,cli
 Author: shbernal

{atsfill-0.0.6.dev203 → atsfill-0.0.6.dev205}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "atsfill"
-version = "0.0.6.dev203"
+version = "0.0.6.dev205"
 description = "Read, dry-run and submit job application forms on hiring systems: Ashby, Greenhouse, Teamtailor, Welcome to the Jungle."
 readme = "README.md"
 license = "MIT"

{atsfill-0.0.6.dev203 → atsfill-0.0.6.dev205}/pyproject.toml.orig

@@ -1,6 +1,6 @@
 [project]
 name = "atsfill"
-version = "0.0.6.dev203"
+version = "0.0.6.dev205"
 description = "Read, dry-run and submit job application forms on hiring systems: Ashby, Greenhouse, Teamtailor, Welcome to the Jungle."
 readme = "README.md"
 license = "MIT"

{atsfill-0.0.6.dev203 → atsfill-0.0.6.dev205}/src/atsfill/adapters/join.py

@@ -48,6 +48,10 @@ SUBMIT_TIMEOUT_MS = 60_000
 BUTTON_TIMEOUT_MS = 2_000
 OPENING_TIMEOUT_S = 20
 CONTINUE_TIMEOUT_S = 15
+ADVANCE_TIMEOUT_MS = 3_000  # how long a clicked card takes to move the wizard on by itself
+INPUT_TIMEOUT_MS = 3_000  # how long a step takes to show its own input
+CLICKS = 3  # a click on a step still rendering is lost, on a card or on Continue
+LEAVE_TIMEOUT_MS = 5_000  # how long Continue takes to leave a step, per click
 POLL_MS = 250
 MAX_STEPS = 60
 MAX_MONTHS = 36  # how far the calendar is paged for a date
@@ -107,6 +111,17 @@ _SHOWN = """root => {
   return p ? p.textContent.trim() : "";
 }"""
 
+# Continue, clicked in one go with the check that the wizard is still on the step: the footer's
+# last button on `review` is Confirm & apply, the send, and a step can leave on its own.
+_CONTINUE = """(button, step) => {
+  const on = location.pathname.replace(/\\/$/, "").split("/").pop() === step;
+  if (!on || button.type === "submit") return false;
+  button.click();
+  return true;
+}"""
+
+_CHECKED = """cards => cards.map(card => card.getAttribute("aria-checked") === "true")"""
+
 # The ids of the inputs that are empty, of those given.
 _EMPTY = (
     """ids => ids.filter(id => { const e = document.getElementById(id); return e && !e.value; })"""
@@ -387,10 +402,10 @@ class Join:
                 if isinstance(answer, str):
                     if answer not in options or options.index(answer) >= len(texts):
                         raise NotTaken(f"no option {answer!r} in {field_id}: {texts}")
-                    card = cards.nth(options.index(answer))
-                    if card.get_attribute("aria-checked") != "true":
-                        card.click()
-                checked = [c.get_attribute("aria-checked") == "true" for c in cards.all()]
+                    if _take(page, cards, options.index(answer)):
+                        self._seen[field_id] = answer  # the wizard took it and moved on
+                        return
+                checked = cards.evaluate_all(_CHECKED)
                 if True in checked and checked.index(True) < len(options):
                     self._seen[field_id] = options[checked.index(True)]
             case "DATE":
@@ -402,7 +417,9 @@ class Join:
             case _ if isinstance(answer, str):
                 name = f"screeningQuestions.answer-{question['id']}"
                 target = page.locator(f'[name="{name}"], [id="{name}"], #salaryExpectationAmount')
-                if not target.count():
+                try:  # the step before's input stays a moment after the URL has changed
+                    target.first.wait_for(timeout=INPUT_TIMEOUT_MS)
+                except Timeout:
                     target = page.locator(
                         "textarea:visible, input:visible:not([type=file]):not([role=combobox])"
                     )
@@ -419,19 +436,41 @@ def _continue(page: Page) -> Locator:
     return page.locator("button:visible:has(> [data-button-label])").last
 
 
+def _take(page: Page, cards: Locator, index: int) -> bool:
+    """Clicks a card until it is checked, and says whether the wizard then moved on by itself."""
+    on = page.url
+    for _ in range(CLICKS):
+        if cards.evaluate_all(_CHECKED)[index]:
+            break
+        cards.nth(index).click()
+        try:
+            page.wait_for_url(lambda url: url != on, timeout=ADVANCE_TIMEOUT_MS)
+        except Timeout:
+            continue
+        return True
+    return False
+
+
 def _next(page: Page, step: str) -> None:
+    """Leaves the step with Continue, unless the wizard has already left it."""
     button = _continue(page)
     deadline = time.monotonic() + CONTINUE_TIMEOUT_S  # an upload enables it once stored
-    while button.is_disabled() and time.monotonic() < deadline:
+    while _step(page) == step and button.is_disabled() and time.monotonic() < deadline:
         page.wait_for_timeout(POLL_MS)
+    if _step(page) != step:
+        return
     if button.is_disabled():
         raise NotTaken(f"the {step} step keeps Continue disabled: a required answer is missing")
     on = page.url
-    try:
-        button.click(timeout=STEP_TIMEOUT_MS)
-        page.wait_for_url(lambda url: url != on, timeout=STEP_TIMEOUT_MS)
-    except Timeout:
-        raise RuntimeError(f"Continue on the {step} step went nowhere") from None
+    for _ in range(CLICKS):
+        try:
+            if not button.evaluate(_CONTINUE, step, timeout=STEP_TIMEOUT_MS):
+                return  # the wizard left the step meanwhile
+            page.wait_for_url(lambda url: url != on, timeout=LEAVE_TIMEOUT_MS)
+        except Timeout:
+            continue
+        return
+    raise RuntimeError(f"Continue on the {step} step went nowhere")
 
 
 def _question(q: dict[str, Any]) -> Field:

{atsfill-0.0.6.dev203 → atsfill-0.0.6.dev205}/src/atsfill/adapters/phenom.py

@@ -43,6 +43,8 @@ JOB = re.compile(r"^(/[a-z]{2,8}/[a-z]{2})/(job|\w*apply)(?:/|$)")
 APPLY = re.compile(r"^/[a-z]{2,8}/[a-z]{2}/\w*apply$")  # `apply`, `hvhapply` for hourly roles
 SEQ = re.compile(r"^/[a-z]{2,8}/[a-z]{2}/job/([A-Za-z0-9]+)")
 CDN = "https://cdn-prod-static.phenompeople.com/CareerConnectResources"
+RENDERER = "MULTI_PAGE_renderer"  # the script of the page that has one, in its loader's data
+MULTI = "multipage"  # what `_phapp_of` adds to `phApp`
 FORM = "form.rjsf"
 NEXT = "button#next"
 TERMS = "terms"  # a notice some sites' wizard opens on, its one button accepting it
@@ -110,6 +112,7 @@ TEMPLATE = re.compile(r"\{\{\s*(.*?)\s*\}\}")
 TAG = re.compile(r"<[^>]+>")
 PAGE_VALUE = re.compile(r"%%(phApp\.[\w.]+)%%")
 ANSWER = re.compile(r"%%formData\.(\w+)%%")
+PLACEHOLDER = re.compile(r"(please )?(choose|select)( (one|an?)\b.*)?$", re.IGNORECASE)
 SEARCHING = re.compile(r"Searching", re.IGNORECASE)  # a typeahead's menu before its results
 ACCEPTED = re.compile(r"\b(DOCX?|PDF|TXT|HTML?|RTF)\b", re.IGNORECASE)
 
@@ -216,12 +219,14 @@ _APPLY_LINK = """seq => {
   return own ? own.href : null;
 }"""
 
-# The errors a step shows, each with its field's control id.
+# The errors a step shows, each with its field's control id and whether the page hides it.
 _ERRORS = """form => [...form.querySelectorAll('.error-detail li, .text-danger, .error-message')]
   .filter(e => e.innerText.trim())
   .map(e => {
     const control = e.closest('.form-group')?.querySelector('input, select, textarea');
-    return [control?.id || control?.name || '?', e.innerText.trim()];
+    const hidden = !!control && !/^(checkbox|radio)$/.test(control.type)
+      && !control.getClientRects().length;
+    return [control?.id || control?.name || '?', e.innerText.trim(), hidden];
   }).slice(0, 4)"""
 
 
@@ -264,6 +269,7 @@ class Phenom:
     known_gaps = {
         "first_step": Gap.FLOW,
         "long_wizard": Gap.FLOW,
+        "hidden_required": Gap.FLOW,
         "calendar_without_month": Gap.RENDERING,
     }
     side_effects = [
@@ -493,6 +499,7 @@ def _phapp_of(html: str) -> dict[str, Any]:
     ph = _literal(html, "var phApp = phApp || ")
     ph["ddo"] = _literal(html, "phApp.ddo = ")
     ph["sessionParams"] = _literal(html, "phApp.sessionParams = ")
+    ph[MULTI] = RENDERER in html
     return ph
 
 
@@ -515,13 +522,11 @@ def _read(
     ph: dict[str, Any], fetch: Callable[[str], Any], post: Post
 ) -> tuple[list[Field], list[dict[str, Any]]]:
     """The fields of every step, from the configuration the page names, and their raw form."""
-    ats = (get(ph, "ddo.response.response.applyATS") or "").lower()
-    base = f"{CDN}/{ph['refNum']}/{ats}/external/careersite/singlepage"
-    config = fetch(f"{base}/schema/form_schema.json")
+    config, bundles = _configuration(ph, fetch)
     texts: dict[str, str] = {}
     for bundle in ("default", ph.get("locale") or "en_us"):
         with suppress(Exception):  # a site has one bundle or both
-            texts |= fetch(f"{base}/bundles/Messages_{bundle}.json")
+            texts |= fetch(f"{bundles}/Messages_{bundle}.json")
     reader = _Config(ph, config, texts, post)
     raws = [raw for step in reader.steps() for raw in step]
     if not raws:
@@ -534,6 +539,27 @@ def _read(
     return ([resume] if resume else []) + [_field(raw) for raw in raws], raws
 
 
+def _configuration(ph: dict[str, Any], fetch: Callable[[str], Any]) -> tuple[dict[str, Any], str]:
+    """The configuration the page's renderer loads, and where its bundles are: a site on the
+    multi-page renderer keeps a stale `singlepage` one, and its form under the page's own name
+    where its `multipage` path has none."""
+    site = f"{CDN}/{ph['refNum']}"
+    ats = (get(ph, "ddo.response.response.applyATS") or "").lower()
+    base = f"{site}/{ats}/external/careersite/{'multipage' if ph.get(MULTI) else 'singlepage'}"
+    try:
+        return fetch(f"{base}/schema/form_schema.json"), f"{base}/bundles"
+    except Exception:
+        if not ph.get(MULTI):
+            raise
+    names = [
+        get(node, "data.page.schemaname")
+        for node in fetch(f"{site}/schema/apply/MULTI_PAGE_form_schema.json")
+        if node.get("type") == "page"
+    ]
+    form = next((name for name in names if name != TERMS), TERMS)
+    return fetch(f"{site}/schema/apply/multipage/{form}.json"), f"{site}/bundles/apply"
+
+
 class _Config:
     """The wizard's steps, each its fields as raw dicts, from the employer's configuration:
     a react-jsonschema-form schema and uiSchema with Phenom's widgets, rules on the job's data,
@@ -703,7 +729,7 @@ class _Config:
                 "datasourcepath": get(u, "ui:options.datasourcepath"),
                 "displayTextField": get(u, "ui:options.dataFieldMapKey") or "label",
             }
-            raw["options"] = _unique(t for _, t in self.loaded_list(listing, api)) or None
+            raw["options"] = _choices(t for _, t in self.loaded_list(listing, api)) or None
         if kind == "select" and re.fullmatch(r"\w+", path):  # a default, the job's country
             raw["default"] = self.prefilled(path)[1] or None
         if kind in ("date", "month", "year"):
@@ -730,7 +756,7 @@ class _Config:
                 scope = path.rpartition(".")[0]
                 at = f"{scope}.{raw['parent']}" if scope else raw["parent"]
                 raw["option_sets"] = [
-                    OptionSet(Is(at, [self.text(parent.get("title"))]), _unique(
+                    OptionSet(Is(at, [self.text(parent.get("title"))]), _choices(
                         self.text(child.get("key"))
                         for child in parent.get("child") or []
                         if child.get("show") is not False
@@ -738,7 +764,7 @@ class _Config:
                     for parent in u.get("data") or []
                 ]  # fmt: skip
                 return None
-            return _unique(
+            return _choices(
                 self.text(child.get("key"))
                 for parent in u.get("data") or []
                 for child in parent.get("child") or []
@@ -750,7 +776,7 @@ class _Config:
                  if isinstance(v, dict) and v.get("htmlType") == "dependent_utm"),
                 None,
             )  # fmt: skip
-            return _unique(self.text(parent.get("title")) for parent in tree or [])
+            return _choices(self.text(parent.get("title")) for parent in tree or [])
         if kind == "radio" and not p.get("enum"):
             return YES_NO
         api = u.get("apiConfig") or get(u, "ui:options.apiConfig")
@@ -777,7 +803,7 @@ class _Config:
         self.listed.setdefault(path.rsplit(".", 1)[-1], listed)  # as a workflow names it
         if raw.get("option_sets"):  # a dependent list's, under its parents' answers before any
             return None
-        return _unique(t for _, t in pairs) or None
+        return _choices(t for _, t in pairs) or None
 
     def loaded_list(self, u: dict, api: dict) -> list[tuple[Any, str]]:
         """A list its own loader serves, by the uiSchema's names for the response's parts."""
@@ -851,7 +877,7 @@ class _Config:
             if isinstance(d, dict)
         ]
         if pairs:
-            options = _unique(t for _, t in pairs if t)
+            options = _choices(t for _, t in pairs if t)
             raw["option_sets"] = [OptionSet(when[0] if len(when) == 1 else AllOf(when), options)]
             raw["parent"] = at
         return pairs
@@ -887,7 +913,7 @@ class _Config:
                     shown = dict(zip(map(str, values), texts, strict=False))
                     if isinstance(data, dict) and shown:
                         raw["option_sets"] = [
-                            OptionSet(Is(at, [shown[key]]), _unique(t for _, t in rows if t))
+                            OptionSet(Is(at, [shown[key]]), _choices(t for _, t in rows if t))
                             for key, rows in pairs.items()
                             if key in shown
                         ]
@@ -1291,8 +1317,10 @@ def _json_text(value: Any) -> str:
     return json.dumps("" if value is None else str(value))[1:-1]
 
 
-def _unique(items: Iterable[str]) -> list[str]:
-    return list(dict.fromkeys(i for i in items if i))
+def _choices(items: Iterable[str]) -> list[str]:
+    """A list's options, once each, without an entry that only asks to pick one: an employer's
+    bundle can label a real value so, which no one means to answer."""
+    return list(dict.fromkeys(i for i in items if i and not PLACEHOLDER.match(i)))
 
 
 def _seq(url: str) -> str | None:
@@ -1388,7 +1416,10 @@ def _next(page: Page, number: int) -> None:
                     raise NotTaken(f"step {number}: {detail}")
                 raise RuntimeError(f"step {number} did not pass: {code}: {detail}")
             if errors := page.locator(FORM).evaluate(_ERRORS) if page.locator(FORM).count() else []:
-                shown = "; ".join(f"{dom}: {message}" for dom, message in errors)
+                if hidden := next((dom for dom, _, hides in errors if hides), None):
+                    detail = f"step {number} requires {hidden}, which the page hides"
+                    raise Unsupported(known_gap(Phenom(), "hidden_required", detail))
+                shown = "; ".join(f"{dom}: {message}" for dom, message, _ in errors)
                 raise RuntimeError(f"step {number} did not pass: {shown}")
         raise RuntimeError(f"step {number} did not pass in time")
     finally: