mark readonly element as interactable (#3845)

skyvern/exceptions.py

@@ -668,6 +668,13 @@ class InputToInvisibleElement(SkyvernException):
         )
 
 
+class InputToReadonlyElement(SkyvernException):
+    def __init__(self, element_id: str):
+        super().__init__(
+            f"The element(id={element_id}) now is readonly. Try to interact with other elements, or try to interact with it later when it's not readonly."
+        )
+
+
 class FailedToParseActionInstruction(SkyvernException):
     def __init__(self, reason: str | None, error_type: str | None):
         super().__init__(

skyvern/webeye/actions/handler.py

@@ -38,6 +38,7 @@ from skyvern.exceptions import (
     IllegitComplete,
     ImaginaryFileUrl,
     InputToInvisibleElement,
+    InputToReadonlyElement,
     InteractWithDisabledElement,
     InteractWithDropdownContainer,
     InvalidElementForTextInput,
@@ -1225,6 +1226,17 @@ async def handle_input_text_action(
     # force to move focus back to the element
     await skyvern_element.get_locator().focus(timeout=timeout)
 
+    # check if the element is readonly(some elements will be non-readonly after focused)
+    if await skyvern_element.is_readonly(dynamic=True):
+        LOG.warning(
+            "Try to input text on a readonly element",
+            task_id=task.task_id,
+            step_id=step.step_id,
+            element_id=skyvern_element.get_id(),
+            action=action,
+        )
+        return [ActionFailure(InputToReadonlyElement(element_id=skyvern_element.get_id()))]
+
     # check the phone number format when type=tel and the text is not a secret value
     if not is_secret_value and await skyvern_element.get_attr("type") == "tel":
         try:
@@ -2939,6 +2951,13 @@ async def select_from_emerging_elements(
         if current_text == actual_value:
             return ActionSuccess()
 
+        if await input_element.is_readonly(dynamic=True):
+            LOG.warning(
+                "Try to input text on a readonly element",
+                element_id=element_id,
+            )
+            return ActionFailure(InputToReadonlyElement(element_id=element_id))
+
         await input_element.input_clear()
         await input_element.input_sequentially(actual_value)
         return ActionSuccess()
@@ -3084,6 +3103,16 @@ async def select_from_dropdown(
                 single_select_result.action_result = ActionSuccess()
                 return single_select_result
 
+            if await input_element.is_readonly(dynamic=True):
+                LOG.warning(
+                    "Try to input text on a readonly element",
+                    element_id=element_id,
+                    task_id=task.task_id,
+                    step_id=step.step_id,
+                )
+                single_select_result.action_result = ActionFailure(InputToReadonlyElement(element_id=element_id))
+                return single_select_result
+
             await input_element.input_clear()
             await input_element.input_sequentially(actual_value)
             single_select_result.action_result = ActionSuccess()

skyvern/webeye/scraper/domUtils.js

@@ -790,13 +790,7 @@ function isInteractableInput(element, hoverStylesMap) {
   // "city", "state", "zip", "country"
   // That's the reason I (Kerem) removed the valid input types check
   var type = element.getAttribute("type")?.toLowerCase().trim() ?? "text";
-  var readOnly = isReadonlyElement(element);
+  return isHoverPointerElement(element, hoverStylesMap) || type !== "hidden";
-
-  return (
-    isHoverPointerElement(element, hoverStylesMap) ||
-    isReadonlyInputDropdown(element) ||
-    (!readOnly && type !== "hidden")
-  );
 }
 
 function isValidCSSSelector(selector) {

skyvern/webeye/scraper/scraper.py

@@ -37,10 +37,11 @@ ScrapeExcludeFunc = Callable[[Page, Frame], Awaitable[bool]]
 RESERVED_ATTRIBUTES = {
     "accept",  # for input file
     "alt",
-    "shape-description",  # for css shape
     "aria-checked",  # for option tag
     "aria-current",
+    "aria-disabled",
     "aria-label",
+    "aria-readonly",
     "aria-required",
     "aria-role",
     "aria-selected",  # for option tag
@@ -48,7 +49,6 @@ RESERVED_ATTRIBUTES = {
     "data-original-title",  # for bootstrap tooltip
     "data-ui",
     "disabled",  # for button
-    "aria-disabled",
     "for",
     "href",  # For a tags
     "maxlength",
@@ -58,6 +58,7 @@ RESERVED_ATTRIBUTES = {
     "readonly",
     "required",
     "selected",  # for option tag
+    "shape-description",  # for css shape
     "src",  # do we need this?
     "text-value",
     "title",
@@ -897,16 +898,23 @@ class IncrementalScrapePage(ElementTreeBuilder):
 
 def _should_keep_unique_id(element: dict) -> bool:
     # case where we shouldn't keep unique_id
-    # 1. not disable attr and no interactable
+    # 1. no readonly attr and not disable attr and no interactable
-    # 2. disable=false and intrecatable=false
+    # 2. readonly=false and disable=false and interactable=false
 
     attributes = element.get("attributes", {})
-    if "disabled" not in attributes and "aria-disabled" not in attributes:
+    if (
+        "disabled" not in attributes
+        and "aria-disabled" not in attributes
+        and "readonly" not in attributes
+        and "aria-readonly" not in attributes
+    ):
         return element.get("interactable", False)
 
     disabled = attributes.get("disabled")
     aria_disabled = attributes.get("aria-disabled")
-    if disabled or aria_disabled:
+    readonly = attributes.get("readonly")
+    aria_readonly = attributes.get("aria-readonly")
+    if disabled or aria_disabled or readonly or aria_readonly:
         return True
     return element.get("interactable", False)
 

skyvern/webeye/utils/dom.py

@@ -281,6 +281,43 @@ class SkyvernElement:
 
         return disabled or aria_disabled or style_disabled
 
+    async def is_readonly(self, dynamic: bool = False) -> bool:
+        # if attr not exist, return None
+        # if attr is like 'readonly', return empty string or True
+        # if attr is like `readonly=false`, return the value
+        readonly = False
+        aria_readonly = False
+
+        readonly_attr: bool | str | None = None
+        aria_readonly_attr: bool | str | None = None
+        mode: typing.Literal["auto", "dynamic"] = "dynamic" if dynamic else "auto"
+
+        try:
+            readonly_attr = await self.get_attr("readonly", mode=mode)
+            aria_readonly_attr = await self.get_attr("aria-readonly", mode=mode)
+        except Exception:
+            LOG.exception(
+                "Failed to get the readonly attribute",
+                element=self.__static_element,
+                element_id=self.get_id(),
+            )
+
+        if readonly_attr is not None:
+            # readonly_attr should be bool or str
+            if isinstance(readonly_attr, bool):
+                readonly = readonly_attr
+            if isinstance(readonly_attr, str):
+                readonly = readonly_attr.lower() != "false"
+
+        if aria_readonly_attr is not None:
+            # aria_readonly_attr should be bool or str
+            if isinstance(aria_readonly_attr, bool):
+                aria_readonly = aria_readonly_attr
+            if isinstance(aria_readonly_attr, str):
+                aria_readonly = aria_readonly_attr.lower() != "false"
+
+        return readonly or aria_readonly
+
     async def is_selectable(self) -> bool:
         return await self.get_selectable() or self.get_tag_name() in SELECTABLE_ELEMENT